From Open Data to Open Science?: A Semantic Diagnosis of Public Science and Technology Data

From Open Data to Open Science?: A Semantic Diagnosis of Public Science and Technology Data | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article From Open Data to Open Science?: A Semantic Diagnosis of Public Science and Technology Data Junyoung Jeong This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8440351/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Open Science has emerged as a central paradigm in contemporary science and technology (S&T) policy, with Open Data widely regarded as one of its core components. Despite this prominence, limited empirical attention has been paid to whether Open Data occupies a structurally meaningful position within the semantic architecture of Open Science discourse. This study conducts a computational semantic analysis of public S&T data-related documents to diagnose the conceptual relationship between Open Data and Open Science. Using BERTopic-based modeling and hierarchical clustering, we examine how Open Data is positioned within the broader Open Science discourse, focusing on its centrality, proximity to key Open Science concepts, and alignment with FAIR principles. The results reveal that while Open Data is frequently referenced, it exhibits a distinct core-periphery structure: administrative and management-oriented metadata occupy a central semantic position, whereas scientifically rich raw data tend to remain on the periphery. The structural analysis further indicates that the semantic integration of Open Data remains uneven across domains, suggesting a partial decoupling between policy expectations and conceptual implementation. By providing a semantic diagnosis of Open Data within Open Science discourse, this study contributes to scientometric research by offering a structural perspective on how foundational concepts of Open Science are articulated and operationalized in practice. The findings highlight the need to move beyond declarative commitments toward a more conceptually integrated understanding of Open Data in the evolution of Open Science. Open Science Open Data Semantic Analysis Scientometrics FAIR Principles Science and Technology Policy Figures Figure 1 Figure 2 Figure 3 Figure 4 1. INTRODUCTION With the acceleration of the Fourth Industrial Revolution, data has become a strategic asset underpinning scientific progress, economic competitiveness, and evidence-based policymaking. Frequently described as the “oil of the 21st century,” data—particularly that generated or held by governments—has been increasingly released as Open Data to enhance transparency, accountability, and innovation (Janssen et al., 2012 ; Bertot et al., 2010 ). International organizations such as the OECD have consistently emphasized that open government data constitutes a foundational infrastructure for data-driven innovation and inclusive growth (OECD, 2015 ; OECD, 2021 ). South Korea represents a leading case in the global Open Data landscape. Following the enactment of the Act on Promotion of the Provision and Use of Public Data , the Korean government rapidly expanded the scale of public data disclosure through its national Open Data Portal (data.go.kr). While this quantitative expansion successfully lowered access barriers and stimulated early-stage data utilization, recent scholarship increasingly questions whether mere availability translates into meaningful reuse and value creation (Zuiderwijk & Janssen, 2014 ; Zuiderwijk et al., 2014 ). As artificial intelligence (AI), machine learning, and large-scale data integration become central to scientific inquiry, qualitative dimensions such as usability, interoperability, and structural consistency have emerged as critical bottlenecks (Borgman, 2015 ; Edwards et al., 2011 ). These challenges are particularly salient in the science and technology (S&T) domain, where public data disclosure is closely intertwined with the broader paradigm of Open Science. Open Science extends beyond open access to publications, advocating transparency across the entire research lifecycle, including data, methods, software, and workflows (Fecher & Friesike, 2014 ; Vicente-Saez & Martinez-Fuentes, 2018 ). From this perspective, publicly funded research data is not merely an administrative byproduct but a core scientific infrastructure enabling reproducibility, cumulative knowledge production, and interdisciplinary convergence (Leonelli, 2016 ; Stodden et al., 2016 ). Within the Open Science discourse, the FAIR Guiding Principles—Findable, Accessible, Interoperable, and Reusable—have become the dominant evaluative framework for assessing data stewardship quality (Wilkinson et al., 2016 ). Subsequent studies have further emphasized that interoperability and reusability are decisive conditions for machine-actionable science, particularly in AI-driven research environments (Jacobsen et al., 2020 ; Mons et al., 2017 ). Data that lacks standardized schemas, controlled vocabularies, or consistent column-level semantics remains difficult to integrate, regardless of its formal openness, effectively functioning as a “data dump” rather than a reusable scientific asset (Murray-Rust, 2008 ; Borgman, 2019a ). Despite the conceptual centrality of FAIR principles, empirical research on public data—especially within the S&T sector—has remained skewed toward policy evaluation, portal-level performance metrics, or user satisfaction surveys (Ubaldi, 2013 ). While recent studies have applied text mining and topic modeling to analyze descriptive metadata such as titles and abstracts (Lee, 2020 ), far less attention has been paid to structural metadata , including column names, variable definitions, and implicit semantic hierarchies. This omission is nontrivial: in data-intensive scientific domains, inconsistencies at the column level directly undermine interoperability, reproducibility, and cross-domain integration (Kučera et al., 2013 ; Munafò et al., 2017 ; Goodman et al., 2016 ). To address this gap, the present study adopts a scientometric and computational semantic perspective to examine how public S&T data is structurally positioned within the discourse of Open Science. Rather than treating Open Data as a static policy output, we conceptualize it as a semantic infrastructure whose internal organization reflects—and potentially constrains—the realization of Open Science principles (Blei & Lafferty, 2006 ; Hecker et al., 2018 ). Methodologically, we employ BERTopic, a state-of-the-art neural topic modeling approach that leverages contextualized embeddings to overcome the sparsity and polysemy limitations of traditional Latent Dirichlet Allocation (Blei et al., 2003 ; Qiang et al., 2020 ; Grootendorst, 2022 ). By extending the analytical scope beyond descriptive text to include key column names and hierarchical topic structures, this study empirically maps the semantic topology of public S&T data. Through dimensionality reduction (UMAP), density-based clustering (HDBSCAN), and diversity-aware keyword extraction (MMR), we investigate whether and how current public data practices support the FAIR principles—particularly interoperability and reusability—at a structural level (Reimers & Gurevych, 2019 ; McInnes et al., 2018 ; Campello et al., 2013 ; Carbonell & Goldstein, 1998 ). Ultimately, this study seeks to bridge the gap between Open Data policy ambitions and their practical realization within Open Science. By revealing the latent semantic structure and positional centrality of S&T data topics, we aim to answer a fundamental question confronting contemporary data governance: Does Open Data, as currently implemented, genuinely function as an enabling infrastructure for Open Science? Accordingly, this study addresses the following research questions: RQ1. What semantic roles does Open Data play within the discourse of Open Science in the science and technology sector? RQ2. To what extent does Open Data occupy a central or peripheral position in the semantic structure of Open Science? RQ3. How does the semantic positioning of Open Data relate to key Open Science principles—particularly the FAIR principles—across domains and over time? 2. Theoretical Background 2.1. Open Data, Metadata, and the Problem of Usability Open Data is broadly defined as data that can be freely accessed, used, modified, and redistributed by any actor without technical or legal restrictions. In the context of public governance, Open Data has been widely conceptualized as a key infrastructural resource for enhancing transparency, accountability, and innovation across sectors (Janssen et al., 2012 ; Bertot et al., 2010 ). International policy organizations, particularly the OECD, have framed Open Data as a foundational component of the digital economy, emphasizing its potential to generate social value, stimulate economic growth, and enable data-driven policy-making (Ubaldi, 2013 ; OECD, 2015 ; OECD, 2021 ). Following the enactment of the Act on Promotion of the Provision and Use of Public Data, South Korea has pursued comprehensive, government-wide open data initiatives, achieving world-leading performance in quantitative indicators such as the number of released datasets and international open data rankings. However, as open data policies have matured, scholarly attention has increasingly shifted from questions of availability toward deeper concerns regarding usability, interpretability, and data quality (Zuiderwijk & Janssen, 2014 ; Zuiderwijk et al., 2014 ). A growing body of literature suggests that the mere release of datasets does not guarantee meaningful reuse. Empirical studies have shown that users frequently encounter difficulties in understanding, integrating, and repurposing open data due to insufficient contextualization and inconsistent data structures (Tenopir et al., 2011 ; Kubler et al., 2018 ). This recognition has repositioned metadata as a central determinant of open data usability and value creation (Kučera et al., 2013 ; Borgman, 2015 ). Metadata—commonly defined as “data about data”—provides critical information regarding the content, structure, provenance, and conditions of use of datasets (Borgman, 2012 ; Mons et al., 2017 ). Prior research has conceptualized metadata as a multi-layered construct encompassing descriptive, structural, and administrative dimensions (Kučera et al., 2013 ; Kubler et al., 2018 ). Descriptive metadata, such as titles, abstracts, and keywords, primarily facilitates dataset discovery and thematic identification. Structural metadata, including schemas, variable definitions, data types, and column names, governs the technical feasibility of machine processing, interoperability, and large-scale integration (Wilkinson et al., 2016 ; Jacobsen et al., 2020 ). Despite this distinction, existing domestic and international studies have predominantly focused on evaluating the completeness and accuracy of descriptive metadata. For example, Lee ( 2020 ) applied text mining techniques to metadata descriptions to identify thematic biases in public data portals, while Kučera et al. ( 2013 ) proposed quantitative indicators to assess missing values and syntactic correctness within metadata fields. In contrast, systematic analyses of structural metadata—particularly the semantic consistency and standardization of column names embedded within raw datasets—remain relatively underdeveloped. The absence of standardized structural metadata substantially increases preprocessing costs, limits cross-domain integration, and undermines the cumulative reuse of data (Zuiderwijk et al., 2014 ; Kubler et al., 2018 ). From a scientometric perspective, such structural fragmentation constrains the formation of scalable knowledge infrastructures and diminishes the long-term scientific and economic value of Open Data (Edwards et al., 2011 ; Borgman, 2019a ). Addressing this gap requires analytical approaches that move beyond surface-level metadata descriptions to diagnose the deeper semantic and structural organization of open data ecosystems. 2.2. Open Science and FAIR-Oriented Data Infrastructures Within the science and technology domain, data disclosure is embedded in the broader paradigm of Open Science. Open Science represents a systemic transformation of the research lifecycle, extending beyond Open Access publications to encompass the sharing of raw data, methodologies, software, and intermediate research outputs (Fecher & Friesike, 2014 ; Vicente-Saez & Martinez-Fuentes, 2018 ). This movement reflects both a normative commitment to science as a public good and a pragmatic response to increasing concerns regarding research transparency and reproducibility (Munafò et al., 2017 ; Stodden et al., 2016 ). Research data occupy a central position in Open Science infrastructures, as they enable verification, replication, and cumulative knowledge production (Leonelli, 2016 ; Goodman et al., 2016 ). Empirical studies have demonstrated that openly shared data are associated with higher citation rates and increased scientific impact, reinforcing the instrumental value of Open Data within scientific ecosystems (Piwowar & Vision, 2013 ; Piwowar et al., 2018 ). The FAIR Guiding Principles—Findable, Accessible, Interoperable, and Reusable—have emerged as the de facto international standard for scientific data stewardship (Wilkinson et al., 2016 ). Unlike earlier openness frameworks, FAIR explicitly emphasizes machine-actionability and interoperability as prerequisites for scalable data reuse (Mons et al., 2017 ; Jacobsen et al., 2020 ). Among the four principles, interoperability is particularly critical, as it requires data to be described using standardized vocabularies, shared schemas, and consistent structural metadata. When datasets generated by different research groups employ harmonized column names, variable definitions, and measurement units, integrated analysis can proceed with minimal friction. Conversely, datasets lacking such structural alignment remain isolated “data silos,” severely limiting their contribution to reproducible and data-intensive science (Leonelli, 2016 ; Murray-Rust, 2008 ). Studies on reproducible research further indicate that access to raw data alone is insufficient if the structural and contextual information necessary to reconstruct analytical processes is absent (Munafò et al., 2017 ; Stodden et al., 2016 ). Despite the centrality of Open Data in Open Science discourse, existing studies have largely focused on policy declarations, compliance indicators, or normative frameworks. Relatively little attention has been paid to how Open Data is semantically positioned within large-scale data infrastructures and whether it functions as a structurally enabling component of Open Science in practice. This gap motivates the need for empirical analyses that examine how Open Data is organized, clustered, and conceptually related to core Open Science principles across domains. 2.3. Computational Topic Modeling and Semantic Analysis Using BERTopic To investigate the latent semantic organization of large-scale textual data, researchers have employed a range of topic modeling techniques. Latent Dirichlet Allocation (LDA) has long served as the dominant approach, modeling documents as probabilistic mixtures of topics based on word co-occurrence patterns (Blei et al., 2003 ). However, LDA’s “bag-of-words” assumption prevents it from capturing contextual meaning and word order, rendering it vulnerable to polysemy and semantic ambiguity (Blei & Lafferty, 2006 ). These limitations are particularly pronounced in the analysis of short texts, such as dataset titles, keywords, and metadata descriptions, where sparsity often leads to incoherent or unstable topics (Qiang et al., 2020 ). To address these challenges, recent studies have increasingly adopted neural embedding-based approaches that preserve contextual semantics. BERTopic represents a state-of-the-art topic modeling framework that integrates transformer-based language models with density-based clustering techniques (Grootendorst, 2022 ). By embedding documents using Sentence-BERT (Reimers & Gurevych, 2019 ), BERTopic captures semantic similarity at the sentence level. Dimensionality reduction is performed using UMAP (McInnes et al., 2018 ), followed by clustering with HDBSCAN, which enables the identification of hierarchical topic structures without requiring a predefined number of topics (Campello et al., 2013 ). Representative keywords are extracted using class-based TF-IDF (c-TF-IDF), with diversity further enhanced through Maximal Marginal Relevance (Carbonell & Goldstein, 1998 ). Because BERTopic clusters documents based on semantic proximity rather than lexical frequency, it is particularly well-suited for analyzing heterogeneous metadata environments where descriptive and structural elements are interwoven. Moreover, its capacity to visualize hierarchical and spatial topic relationships aligns with the objectives of scientometric research that seeks to uncover both micro-level thematic patterns and macro-level structural configurations. Accordingly, this study employs BERTopic to empirically diagnose the semantic topology of public science and technology Open Data and to assess its structural alignment with the principles of Open Science. 3. MATERIALS AND METHODS 3.1. Data Collection and Preprocessing To empirically analyze the semantic structure and standardization status of Open Data in the domestic science and technology sector, this study selected the "Public Data Registration Management System," operated by the Ministry of the Interior and Safety, as the primary source for data collection. This system serves as a government-wide platform that registers and manages metadata for data held by central ministries, local governments, and public institutions, providing an optimal sample for identifying the status of national data initiatives. Data collection was conducted as of December 23, 2024, targeting all file data registered under the "Science and Technology" category within the system's classification scheme. Although a total of 4,562 data entries were identified, the raw data contained significant noise and a high volume of entries with low metadata fidelity, rendering them unsuitable for topic modeling. Consequently, a stepwise data cleaning process was implemented to ensure analysis reliability and optimize the performance of the BERTopic model. First, De-identification of Institution Names was performed. Public data titles frequently include the names of providing institutions, such as "[Ministry of Environment]," "(Foundation) Research Institute," or "OO Center." Since these names are irrelevant to the intrinsic subject matter and can induce bias where data clusters around "specific institutions" rather than topics, we utilized Regular Expressions to systematically remove text within square brackets [] and parentheses () along with the enclosed institution names, retaining only the pure keywords. Second, Normalization and Concatenation were conducted. After removing unnecessary special characters and excessive spacing, three metadata fields—1) Title, 2) Description, and 3) Keywords & Column Names—were concatenated into a single document. This step was taken to mitigate the sparsity problem inherent in short texts and to enable the BERTopic model to learn the semantic context more robustly. Third, Morphological Analysis was performed. Given the agglutinative nature of the Korean language, tokenization based simply on spacing has limitations in capturing meaning. Therefore, we utilized the Okt (Open Korean Text) morphological analyzer to extract nouns, which are the core parts of speech constituting the meaning of sentences. Through this rigorous preprocessing pipeline, 1,070 data entries were finally confirmed as the valid analysis corpus. This corresponds to approximately 23.5% of the initially collected data, a selection made to enhance the precision of the analysis results by utilizing only high-quality text data from which noise had been eliminated. 3.2. BERTopic Methodology This study utilized the BERTopic algorithm to extract latent topics and capture their nuanced context from the collected unstructured text data. BERTopic is a sophisticated model proposed to overcome the "Bag-of-Words" limitations inherent in traditional probabilistic models like Latent Dirichlet Allocation (LDA), which typically ignores word order and broader semantic context. By combining pre-trained Transformer-based embeddings with a class-based TF-IDF (c-TF-IDF) procedure, the model maximizes the semantic consistency and coherence of the derived topics. Regarding the specific analysis procedure, we first applied the SBERT (Sentence-BERT) framework to transform each document into a dense, high-dimensional vector space. To accurately capture the intricate contextual nuances of Korean text, we employed the paraphrase-multilingual-MiniLM-L12-v2 model. This specific model is specialized for multilingual processing and is designed to position semantically similar sentences closely within the vector space, thereby mitigating the data sparsity issues common in short-text metadata. Since the generated embedding vectors possess high dimensionality (over 384 dimensions), the UMAP (Uniform Manifold Approximation and Projection) algorithm was performed to prevent the "curse of dimensionality" that often occurs during density-based clustering. UMAP is a non-linear dimensionality reduction technique that effectively reduces dimensions while preserving both the local and global topological structure of the data. In this study, to preserve semantic similarity for subsequent clustering, dimensions were reduced to 5 components using cosine distance, with the number of neighbors (n_neighbors) set to 15 to balance the focus between local and global structures. For clustering the dimensionally reduced vectors, HDBSCAN (Hierarchical Density-Based Spatial Clustering of Applications with Noise) was applied. HDBSCAN offers a significant advantage over partition-based methods (like K-Means) as it does not require a prior specification of the number of clusters and enhances topic clarity by treating low-density outliers as noise. To prevent the generation of excessively fragmented or microscopic topics, the minimum cluster size (min_cluster_size) was set to 10. Finally, the keywords representing each clustered topic were extracted using the c-TF-IDF method. This method calculates word importance by considering clusters, rather than individual documents, as the unit of analysis. 3.3. Research Procedure This study established and executed a systematic three-stage framework, as shown in Fig. 1 , to identify the semantic structure of public data in the science and technology sector and empirically analyze how it connects to the actual Open Science ecosystem. The first stage, Data Collection, involves the selection and acquisition of raw datasets in the science and technology field, which serve as the subject of analysis. The data collected in this process goes beyond physical aggregation; it is checked for standardization status and refined into an analyzable form to serve as the foundational material for subsequent analysis. The second stage is Topic Modeling, which begins with a preprocessing step to improve the quality of the collected unstructured text data. After stop-word removal and morphological analysis, the text data is input into the BERTopic algorithm, the core methodology of this study. Through a sequence of embedding, dimensionality reduction, and clustering, latent topics within the text are derived, thereby concretizing the semantic structure of the science and technology data. The final third stage is Topic Analysis. For each topic generated through modeling, Topic Labeling is performed by reviewing key keywords and representative documents to assign subject names. This is followed by a Topic Classification process that categorizes themes of a similar nature. Finally, the study derives its conclusions by performing an Evaluation of Linkage to Open Science, which analyzes how well these structured topics align with the core values of Open Science—openness, sharing, and collaboration—and determines whether current data disclosure leads to the revitalization of Open Science. Figure 2. Research procedure framework: From data collection to semantic analysis 4. RESULTS 4.1. Topic Modeling Analysis This study applied BERTopic modeling (Grootendorst, 2022 ) to public science and technology (S&T) datasets to uncover latent semantic structures embedded in large-scale metadata. Compared to probabilistic topic modeling approaches such as Latent Dirichlet Allocation (LDA), which are known to suffer from sparsity and semantic dilution when applied to short and heterogeneous metadata texts (Blei et al., 2003 ; Qiang et al., 2020 ), BERTopic leverages contextual sentence embeddings to capture semantic similarity at a higher conceptual resolution. Recent scientometric studies have emphasized that embedding-based topic modeling is particularly suitable for policy-oriented and infrastructure-related corpora, where terminological variation is high and semantic nuance is critical (Dai, 2025 ; Walker et al., 2019 ). The final model achieved a silhouette coefficient of 0.681, indicating strong topic separability and internal coherence. This level of clustering performance is comparable to or exceeds benchmarks reported in recent large-scale semantic analyses of research policy documents and public data repositories (Gokhberg et al., 2023 ; Aria et al., 2020 ). After excluding noise documents identified through density-based clustering, a total of 30 coherent topics were extracted, providing a granular semantic representation of the national public S&T data landscape. A closer examination of topic keywords reveals that Topic 0 (Intellectual Property & Patent Status) occupies a structurally prominent position, characterized by high-frequency terms such as registration number , application number , and publication number . This dominance suggests that a substantial portion of public S&T data is oriented toward the legal and institutional management of research outputs. Prior studies have noted that intellectual property–related metadata often functions as an interface between public research systems and market-oriented innovation ecosystems, reinforcing the economic rationales underlying open data policies (Kitchin, 2014 ; Ruijer & Meijer, 2020 ). Similarly, topics related to academic communication—such as Topic 7 (Academic Literature & Bibliographic Info), Topic 18 (Theses & Research Publications), and Topic 19 (Digital Content Metadata)—are defined by descriptors including author , paper title , and attachment link . These topics confirm that traditional scholarly outputs remain a foundational component of public data infrastructures, consistent with earlier findings that open bibliographic metadata serves as the backbone of knowledge dissemination and scientometric analysis (Borgman, 2012 ; Waltman, 2016 ). From the perspective of research reproducibility, the identification of multiple topics containing raw observational and experimental data is particularly noteworthy. Topics such as Topic 16 (Weather & Environment Observation), Topic 23 (Bio-health & Body Composition), and Topic 4 (Network Quality & Speed) include concrete measurement variables (e.g., wind speed , body fat mass , Mbps ), which are essential for independent verification and secondary analysis. This finding aligns with recent empirical work emphasizing that the availability of machine-readable raw data, rather than summarized results alone, is a prerequisite for cumulative and reproducible science (Stodden et al., 2016 ; Goodman et al., 2016 ; Fanelli et al., 2018 ). Topics related to research infrastructure and experimental context—such as Topic 6 (Research Equipment & Specifications) and Topic 13 (Manufacturing Process & Part Data)—contain detailed attributes including model name , manufacturer , and process code . These structural descriptors play a critical role in documenting the conditions under which data are generated, thereby supporting interpretability and reuse. Recent Open Science literature has increasingly highlighted the importance of such contextual metadata as a bridge between openness and actual scientific usability (Edwards et al., 2011 ; Leonelli, 2016 ; Pasquetto et al., 2019 ). Administrative and managerial dimensions of the national R&D system are also strongly represented. Topics such as Topic 3 (ICT R&D Projects & SW Development), Topic 15 (Lab & Project Execution Info), and Topic 21 (Institutional Research Performance Indicators) include identifiers related to project lifecycle management, performance evaluation, and institutional accountability. These findings support the view that public S&T data functions not only as a scientific resource but also as a governance instrument that enhances transparency and coordination within complex research systems (Bertot et al., 2010 ; Janssen et al., 2012 ; Kitchin, 2014 ). At the same time, the emergence of Topic 22 and Topic 28—both characterized exclusively by the absence of identifiable column information—reveals persistent structural weaknesses in metadata provision. The presence of such topics corroborates prior assessments that incomplete or non-standardized structural metadata constitutes a major bottleneck for interoperability and reuse (Kubler et al., 2018 ; Kučera et al., 2013 ). Recent studies further argue that these deficiencies limit the machine-actionability of open data, thereby constraining its integration into advanced analytical pipelines and AI-driven research environments (Wilkinson et al., 2016 ; Mons et al., 2017 ; Jacobsen et al., 2020 ). Overall, the 30 extracted topics can be interpreted as reflecting four interrelated pillars of the public S&T ecosystem: R&D activities, research infrastructure, industry and innovation support, and socio-cultural services. The semantic diversity and concreteness of the identified keywords suggest that current data opening practices are moving beyond symbolic disclosure toward more operational forms of openness. This structural heterogeneity mirrors contemporary conceptualizations of Open Science as a multifaceted system encompassing epistemic, institutional, and infrastructural dimensions (Fecher & Friesike, 2014 ; Vicente-Saez & Martinez-Fuentes, 2018 ). Table 1 List of Derived Topics and Top 5 Keywords Topic Topic Label Top 5 Keywords Topic 0 Intellectual Property & Patent Status Registration number, Application number, Registration date, Title, Publication number Topic 1 Location-based Facilities/Business Info Phone number, Address, Longitude, Latitude, Business name Topic 2 Forest & Biological Resources Info Korea Forest Service, Title, Scientific name, WonPyongOh1993, Common name Topic 3 ICT R&D Projects & SW Development Project number, Participating agency, Participation type, Project name, Classification code Topic 4 Network Quality & Speed Download, Upload, mbps, Carrier, bit Topic 5 Postal Service & Reference Data Data reference date, October, November, December, January Topic 6 Research Equipment & Specifications Model name, Sequence number, Equipment name, Manufacturer, Acquisition date Topic 7 Academic Literature & Bibliographic Info Author, Signature, Paper title, Title, Publication year Topic 8 System Log & Education History Registration date/time, Usage status, Business number, Education name, Institution name Topic 9 Sci-Tech Human Resource Training Training project, Participation, Graduate student, Future talent, Institution name Topic 10 Online Board Activity Status Serial number, Registration date, Views, Title, url Topic 11 Broadcast Ad Market Statistics Year, Type, Ad type, Sales, Million KRW Topic 12 Tech Transfer & Corporate Support Cumulative, Paid, Company, Free, By target Topic 13 Manufacturing Process & Part Data Registration date, Substrate size, Equipment name, Process name, Process code Topic 14 Admin Contact & Civil Service Manager, Department, Project name, Support details, Contact info Topic 15 Lab & Project Execution Info Laboratory, Project classification, Lab name, Researcher name, Location Topic 16 Weather & Environment Observation Wind speed, Wind direction, edr, 50m, 76m Topic 17 Research Agencies & Duration Institution, Execution period, Data title, Data type, File name Topic 18 Theses & Research Publications Title, Sequence number, Address, Thesis title, Language Topic 19 Digital Content Metadata Content, Title, Views, Attachment link, Link Topic 20 Public Facility Location Info Longitude, Latitude, Facility type, Installation place name, Address Topic 21 Institutional Research Performance Indicators Institution name, Type, Data name, Link, Region Topic 22 Unidentifiable Data Absence of column info (Single keyword) Topic 23 Bio-health & Body Composition Water, Extracellular water total water ratio, Item, Body fat mass, Max Topic 24 R&D Workforce Demographics Female, Male, Ratio, PhD, Master Topic 25 Data Lifecycle History Deletion status, Technology overview, Link, Future issue, Classification Topic 26 Radio Management & Station Permit Permit, Management office, Classification code, Gangneung Radio Management Office, MSIT Topic 27 Broadcast Programming & Channel Info Broadcast field, Program name, Broadcaster name, Channel number, Nationality Topic 28 Unidentifiable Data Absence of column info (Single keyword) Topic 29 Broadcast Content Genre Classification Ratio, Education, News/Current affairs, Documentary, Drama 4.2. Thematic Classification and Semantic Structure of Science and Technology Data The resulting dendrogram (Fig. 3 ) reveals that public S&T data do not exist as isolated informational units, but rather form an interconnected semantic ecosystem. This ecosystem exhibits a cyclic topology structured around three overarching dimensions: (1) R&D execution and knowledge production, (2) industrial application and innovation diffusion, and (3) social dissemination and public engagement. Such a configuration empirically supports the conceptualization of Open Science as a systemic process rather than a collection of independent openness practices (Kitchin, 2014 ; Vicente-Saez & Martinez-Fuentes, 2018 ). First, the R&D lifecycle constitutes the semantic core of the ecosystem. Cluster 2 (R&D Project & Performance Management) aggregates administrative and operational data related to project execution, funding duration, and institutional responsibility. Its close semantic linkage with Cluster 1 (IP & Research Outcomes) demonstrates that research outputs—such as patents, publications, and theses—are structurally embedded within administrative data flows. This finding reinforces prior evidence that transparency in R&D management data is critical for mitigating redundant public investment and enhancing accountability in national research systems (Ruijer & Meijer, 2020 ; Ubaldi, 2013 ). Furthermore, the adjacency of Cluster 3 (Research Infrastructure & Standardization) highlights the infrastructural preconditions for reproducible research. The inclusion of equipment specifications, manufacturing processes, and biological resource metadata indicates that openness is extending beyond final outputs toward the disclosure of experimental and procedural contexts, a key requirement for computational and empirical reproducibility (Stodden et al., 2016 ; Goodman et al., 2016 ). Second, the industry and innovation diffusion domain emerges as a distinct yet tightly coupled layer. Cluster 4 (Industry Support & Technology Transfer) captures the semantic intersection between public R&D and private-sector utilization, encompassing technology transfer outcomes, corporate support metrics, and data lifecycle histories. This cluster operationalizes the notion of data-driven innovation, wherein public sector information acts as a foundational input for economic value creation (OECD, 2015 ; Ruijer & Meijer, 2020 ). Cluster 7 (ICT & Communication Infrastructure), positioned as a supporting layer, underscores the role of network quality, radio management, and digital connectivity as enabling conditions for both industrial innovation and large-scale data reuse. Prior studies have emphasized that without interoperable and accessible digital infrastructure, the economic potential of open data remains largely unrealized (Kubler et al., 2018 ; Zuiderwijk et al., 2014 ). Third, the presence of basic science and raw observational data signifies a qualitative maturation of public S&T data openness. Cluster 6 (Basic Science & Observational Data) includes meteorological measurements, environmental observations, and bio-health variables. Unlike aggregated statistical indicators, these datasets retain high analytical granularity, allowing downstream users to repurpose data for novel scientific inquiries. This characteristic directly corresponds to the Reusability dimension of the FAIR principles and reflects a shift from symbolic openness toward functional openness (Wilkinson et al., 2016 ; Pasquetto et al., 2019 ). Empirical research has shown that access to raw data substantially increases citation impact and accelerates cumulative knowledge production (Piwowar & Vision, 2013 ; Fanelli et al., 2018 ). Finally, the social diffusion and public interface layer expands the scope of S&T data beneficiaries beyond expert communities. Cluster 5 (Sci-Tech Human Resource Training) and Cluster 9 (Administrative & Civil Services) represent human capital development and public-facing administrative information, while Cluster 10 (Regional & Location-based Information) spatially anchors scientific resources within local contexts. The emergence of Cluster 8 (Broadcasting & Digital Content) further illustrates the convergence of S&T data with media ecosystems, transforming scientific information into consumable cultural content. This multi-layered structure substantiates arguments that Open Science increasingly functions as an infrastructure for citizen engagement, policy learning, and societal sensemaking (Hecker et al., 2018 ; Pasquetto et al., 2019 ). In sum, the ten Macro-Topics identified through hierarchical clustering empirically demonstrate a virtuous semantic cycle in public S&T data. Research resources and infrastructures (Clusters 3 and 6) are mobilized through R&D activities (Clusters 2 and 5), generating formal outputs (Cluster 1) that diffuse into industrial (Clusters 4 and 7) and societal domains (Clusters 8–10). This structural topology provides concrete evidence that semantic organization plays a pivotal role in determining whether open data can effectively evolve into Open Science. Table 2 Hierarchical Clustering Results: Macro-Topics and Constituent Micro-Topics Cluster Macro-Label Micro-Topics (Topic ID & Label) 1 IP & Research Outcomes Topic 0 (Intellectual Property & Patent Status) Topic 7 (Academic Literature & Bibliographic Info) Topic 18 (Theses & Research Publications) 2 R&D Project & Performance Management Topic 3 (ICT R&D Projects & SW Development) Topic 15 (Lab & Project Execution Info) Topic 17 (Research Agencies & Duration) 3 Research Infrastructure & Standardization Topic 6 (Research Equipment & Specifications) Topic 13 (Manufacturing Process & Part Data) Topic 2 (Forest & Biological Resources Info) 4 Industry Support & Tech Transfer Topic 11 (Broadcast Ad Market Statistics) Topic 12 (Tech Transfer & Corporate Support) Topic 25 (Data Lifecycle History) 5 Sci-Tech Human Resource Training Topic 9 (Sci-Tech Human Resource Training) Topic 24 (R&D Workforce Demographics) 6 Basic Science & Observational Data Topic 16 (Weather & Environment Observation) Topic 23 (Bio-health & Body Composition) Topic 5 (Postal Service & Reference Data) 7 ICT & Comm. Infrastructure Topic 4 (Network Quality & Speed) Topic 26 (Radio Management & Station Permit) 8 Broadcasting & Digital Content Topic 27 (Broadcast Programming & Channel Info) Topic 29 (Broadcast Content Genre Classification) Topic 19 (Digital Content Metadata) 9 Admin & Civil Services Topic 8 (System Log & Education History) Topic 10 (Online Board Activity Status) Topic 14 (Admin Contact & Civil Service) 10 Regional & Location-based Info Topic 1 (Location-based Facilities/Business Info) Topic 20 (Public Facility Location Info) Topic 21 (Institutional Research Performance Indicators) 4.3. Semantic Topology and Spatial Distribution of Topics Figure 4 illustrates the semantic topology and spatial distribution of representative topic clusters derived from public science and technology (S&T) data metadata. The two-dimensional embedding, generated using BERTopic and projected via UMAP, represents latent semantic relationships among datasets, where spatial proximity reflects higher semantic similarity. As emphasized in prior scientometric studies, such low-dimensional embeddings should not be interpreted as literal geometric distances but as relational representations that reveal the underlying conceptual organization of large textual corpora (Börner et al., 2003 ; Gokhberg et al., 2023 ). The overall distribution reveals a densely populated central region primarily composed of clusters related to R&D project management, administrative coordination, and performance evaluation. This finding suggests that administrative and managerial metadata function as a semantic backbone within the public S&T data ecosystem. Similar patterns have been observed in science policy corpora, where governance-related concepts often occupy structurally central positions due to their integrative role across heterogeneous knowledge domains (Aria et al., 2020 ; Waltman, 2016 ). From a FAIR perspective, these centrally located clusters are strongly associated with Findability and Accessibility, as they frequently include persistent identifiers, project codes, institutional affiliations, and access-related attributes that facilitate data discovery and coordination (Wilkinson et al., 2016 ; Jacobsen et al., 2020 ). In contrast, clusters positioned toward the periphery of the embedding space exhibit more domain-specific semantic characteristics. Clusters associated with basic science and observational data—such as meteorological measurements and bio-health datasets—form relatively compact and distinct groupings. These clusters are characterized by the prevalence of raw variables, measurement units, and experimental context descriptors, indicating a stronger alignment with the Reusability dimension of FAIR. This pattern is consistent with empirical findings that raw scientific data tends to be semantically cohesive yet operationally isolated when structural metadata and common vocabularies are insufficiently standardized (Edwards et al., 2011 ; Pasquetto et al., 2019 ). Clusters related to research infrastructure, experimental environments, and standardization practices occupy intermediate positions between central administrative clusters and peripheral raw data clusters. This spatial configuration reflects their dual semantic role. On the one hand, these topics connect to managerial metadata through standardized classifications, process codes, and equipment identifiers. On the other hand, they support experimental reproducibility by documenting physical research settings and technical specifications. Prior studies on reproducibility and data stewardship emphasize that such infrastructural metadata constitutes a critical connective layer enabling cross-domain interoperability (Stodden et al., 2016 ; Leonelli, 2016 ). Meanwhile, clusters associated with industrial support, technology transfer, and digital content appear in localized regions at moderate semantic distances from core research clusters. This pattern suggests that industry-facing datasets are conceptually adjacent to research outputs but remain partially decoupled from raw scientific data. This observation aligns with interpretations of Open Science as an extended ecosystem that includes policy and industry actors, while also highlighting persistent structural boundaries that limit seamless data reuse across institutional sectors (Ruijer & Meijer, 2020 ; OECD, 2021 ). Taken together, the semantic topology depicted in Fig. 4 demonstrates that FAIR-related attributes are unevenly distributed across topic clusters rather than uniformly embedded throughout the public S&T data landscape. The spatial configuration indicates that openness alone does not guarantee interoperability or reusability. Instead, these qualities emerge from the presence of standardized, machine-readable structural metadata that bridge semantically distant domains. This finding reinforces recent arguments in scientometrics and data studies that advancing Open Science requires not only expanding data disclosure but also strategically strengthening metadata infrastructures to enable large-scale integration, reuse, and cumulative knowledge production (Borgman, 2019b ; Mons et al., 2017 ). 4.4. Semantic Diagnosis of Open Science Connectivity based on FAIR Principles To assess whether the semantic structures of the 30 derived topics contribute to the realization of an actual Open Science ecosystem, this study employed the FAIR principles proposed by Wilkinson et al. ( 2016 ) as an analytical framework. The FAIR principles—Findability, Accessibility, Interoperability, and Reusability—serve as essential prerequisites for data to function as scientific assets beyond the scope of mere information disclosure. The results of mapping the keywords and semantic structures extracted via topic modeling to these four criteria are as follows. First, in terms of Findability, it was confirmed that public science and technology data encompass unique identifiers and rich metadata. In Topic 1 (Intellectual Property & Patent Status) and Topic 8 (Academic Literature & Bibliographic Info), unique identifiers such as 'Registration number,' 'Application number,' 'DOI,' 'Author,' and 'Signature' were derived as core keywords. This implies that data can be identified as unique objects, suggesting that an indexing foundation is established to allow researchers to precisely search for necessary resources amidst a vast sea of data. Furthermore, Topic 21 (Public Facility Location Info) includes spatial coordinate data like 'Latitude' and 'Longitude,' thereby extending the findability of GIS-based data into physical space. Second, regarding Accessibility, the data present clear access protocols and usage conditions. Keywords such as 'URL,' 'Link,' and 'Attachment file' appearing in Topic 11 (Online Board Activity Status) and Topic 19 (Digital Content Metadata) indicate that physical paths for users to access the actual data entities are explicitly specified within the metadata. Notably, keywords such as 'Paid,' 'Free,' and 'By target' included in Topic 13 (Tech Transfer & Corporate Support) reveal that information regarding data or technology usage rights and licensing is provided simultaneously. This proves that the opened data exists in a state where it can be actually acquired without legal or economic barriers, or under clearly stipulated conditions. Third, concerning Interoperability, the adoption of machine-readable standard vocabularies and code systems was identified. The keywords 'Scientific name' and 'Common name' derived from Topic 3 (Biodiversity & Forest Resources) demonstrate adherence to international standard classification systems in biology, implying that biological data produced by different research institutions can be integrated based on species information. Additionally, the 'Process code' in Topic 14 (Manufacturing Process & Part Data) and the 'Classification code' in Topic 27 (Radio Management & Station Permit) corroborate that data are managed not as unstructured text but as standardized code values. Such compliance with standards is a critical element enabling data exchange and convergent analysis between heterogeneous systems (Janssen et al., 2012 ). However, the topics identified as 'Absence of column info' in Topic 23 and Topic 29 reveal that some datasets remain in a non-standardized state. This serves as a counter-example highlighting the urgency of advancing data quality management systems. Fourth, this study identified the most significant correlation with Open Science in the aspect of Reusability. Reusability is a core value determining whether data can be utilized as material for new research beyond its initial collection purpose (Murray-Rust, 2008 ). The analysis revealed that Topic 17 (Weather Observation & Aviation Info) and Topic 24 (Bio-health & Body Composition) contain observational variables themselves, such as 'Wind speed,' 'Wind direction,' 'Body fat mass,' and 'Water ratio,' rather than processed statistics. Furthermore, Topic 7 (Research Equipment & Specifications) and Topic 14 (Manufacturing Process & Part Data) provide detailed information on the physical environment and context in which research was conducted, including 'Model name,' 'Manufacturer,' 'Substrate size,' and 'Process name.' This supports the ideals of Open Science—specifically research transparency and reproducibility—by providing a foundation for third-party researchers to verify preceding studies or replicate experiments by reconstructing the same environment using the data (Munafò et al., 2017 ). In conclusion, public science and technology data are evaluated to satisfy the four requirements of the FAIR principles at a semantic level. In particular, the structural identification of raw data and experimental context information, transcending administrative result reports, empirically demonstrates that the current data opening policy is evolving beyond simple transparency enhancement into an Open Science ecosystem for practical scientific value creation (OECD, 2015 ; Vicente-Saez & Martinez-Fuentes, 2018 ). Table 3 demonstrates that different semantic topic clusters align with distinct FAIR principles based on concrete metadata attributes rather than normative assumptions. Table 3 Empirical mapping of representative semantic topics to FAIR principles in public S&T data FAIR Principle Core Requirement Matched Topics (Topic ID & Label) Key Empirical Evidence Findable Persistent identifiers and rich descriptive metadata Topic 1 (Intellectual Property); Topic 8 (Bibliographic Information); Topic 21 (Institutional Indicators) • Persistent identifiers: registration number, application number, DOI • Structured bibliographic metadata enabling systematic discovery Accessible Clear access protocols and usage conditions Topic 11 (Advertising Market Statistics); Topic 19 (Digital Content); Topic 13 (Technology Transfer) • Access paths: URL, hyperlinks, attached files • Explicit usage conditions: free, paid, target-specific licensing Interoperable Standardized vocabularies and machine-readable structures Topic 3 (ICT R&D and Bio-resources); Topic 14 (Administrative Contact and Manufacturing Process); Topic 27 (Broadcast Programming Data) • Controlled vocabularies: scientific names, common names based on bio-standards • Standard classification and process codes enabling cross-dataset integration Reusable Provision of raw data and contextual metadata Topic 17 (Research Agencies and Weather Data); Topic 24 (R&D Workforce and Bio-health); Topic 7 (Research Equipment); Topic 13 (Manufacturing Process Data) • Raw variables: wind speed, wind direction, body fat mass, water ratio • Contextual metadata: model name, manufacturer, substrate size, experimental conditions 5. DISCUSSION This study contributes to the growing body of Open Science and open data research by empirically examining the semantic and structural characteristics of public science and technology (S&T) data through large-scale metadata analysis. Unlike prior studies that primarily evaluate policy compliance or portal-level performance indicators, this research focuses on the latent semantic architecture embedded within unstructured metadata and its alignment with the FAIR principles. By employing BERTopic-based topic modeling and hierarchical clustering, the analysis reveals how public S&T data are conceptually organized across the R&D lifecycle and how structural metadata mediates their scientific usability. The findings demonstrate that public S&T data are semantically structured as an interconnected ecosystem rather than a collection of isolated datasets. The emergence of multiple topic clusters spanning research outcomes, infrastructure, industrial support, and raw observational data suggests that openness has increasingly extended upstream and downstream within the research process. This observation aligns with recent conceptualizations of Open Science as an infrastructural transformation rather than a narrow dissemination practice (Borgman, 2019a ; Edwards et al., 2011 ). In particular, Edwards et al. ( 2011 ) emphasize that data infrastructures shape what kinds of scientific practices become scalable, a perspective that helps explain why administrative and management-related metadata emerge as structurally central in the semantic topology observed in this study. From a FAIR perspective, the semantic proximity between raw data topics and contextual metadata—such as equipment specifications, experimental environments, and process descriptions—indicates a growing orientation toward reproducibility and reuse. This supports recent arguments that reproducibility is not achieved through data availability alone, but through the availability of sufficiently rich contextual information that allows analytical reconstruction (Stodden et al., 2016 ; Goodman et al., 2016 ). The presence of semantically dense clusters related to meteorological and bio-health data further highlights the role of public S&T data as reusable scientific inputs, echoing empirical findings that well-curated datasets generate cumulative value through secondary analysis and cross-domain recombination (Piwowar et al., 2018 ). At the same time, the results expose a critical structural imbalance. While semantically rich topics exist, several clusters exhibit ambiguous or weakly defined structural metadata, limiting their machine-actionable interoperability. This finding resonates with recent critiques that many open data initiatives remain “human-readable but machine-hostile,” thereby constraining large-scale computational reuse (Wilkinson et al., 2016 ; Jacobsen et al., 2020 ). Jacobsen et al. ( 2020 ) argue that without consistent schema definitions and standardized variable-level metadata, FAIR principles cannot be operationalized in practice—a claim strongly supported by the unidentified and structurally fragile topic clusters identified in this analysis. Beyond academic reuse, the semantic coupling between intellectual property, technology transfer, and industrial support topics underscores the innovation-oriented dimension of public S&T data. This configuration suggests that open research data function as boundary objects connecting public research and private-sector innovation systems. Recent studies on data-driven innovation emphasize that such spillovers depend less on openness per se than on interoperability and legal clarity (OECD, 2021 ). In this context, the study’s findings imply that deficiencies in structural metadata may inhibit not only scientific reuse but also downstream industrial valorization. Methodologically, this study advances scientometric and open data research by demonstrating the analytical value of topic modeling grounded in semantic embeddings rather than frequency-based indicators alone. While previous studies have applied topic modeling to identify thematic orientations in public data portals, they often stop short of linking semantic patterns to infrastructural operability (Cho, 2023 ). By integrating semantic clustering with FAIR-oriented interpretation, this research bridges the gap between computational text analysis and data governance theory. Overall, the findings suggest that future Open Science policies should shift emphasis from the quantitative expansion of datasets toward the qualitative strengthening of structural metadata. Mandating standardized, machine-readable schemas, persistent identifiers at the variable level, and explicit contextual documentation would significantly enhance interoperability and reusability. Such measures are increasingly recognized as prerequisites for sustainable Open Science ecosystems capable of supporting reproducibility, cross-sector collaboration, and cumulative knowledge production (Thibault et al., 2023 ). Despite these contributions, this study is subject to several limitations. First, the empirical analysis was restricted to the South Korean Open Data Portal, which may limit the generalizability of the findings to other national contexts with different data governance models. Second, the semantic diagnosis relied primarily on textual and structural metadata without physically validating the internal quality or completeness of the actual data files (e.g., cell-level values), leaving the technical assessment of data integrity for future research. Finally, because the analysis focused on supply-side metadata provided by government agencies, it does not directly capture the demand-side user experience or the actual downstream reuse patterns of the datasets. Table 4 Summary of key empirical findings in relation to research questions Research Question Analytical Focus Key Results RQ1. What semantic roles does Open Data play within the discourse of Open Science in the S&T sector? Semantic clustering and topic interpretation Public S&T data form a multi-layered semantic structure spanning the entire R&D lifecycle, indicating that Open Data functions as a connective infrastructure rather than a collection of isolated datasets. RQ2. To what extent does Open Data occupy a central or peripheral position in the semantic structure of Open Science? Spatial topology and semantic centrality Topics related to R&D project management and administrative metadata occupy central positions, acting as semantic backbones, while scientifically rich raw data topics remain peripheral, revealing structural asymmetries in data integration. RQ3. How does the semantic positioning of Open Data relate to FAIR principles across domains? FAIR-oriented semantic mapping FAIR-aligned attributes are unevenly distributed across topics; interoperability and reusability depend on the presence of standardized structural metadata rather than openness alone. 6. CONCLUSION This study set out to examine how public science and technology (S&T) data are semantically structured and positioned within the evolving Open Science ecosystem, moving beyond conventional evaluations centered on policy compliance or mere data availability. By applying BERTopic-based topic modeling and hierarchical clustering to large-scale unstructured metadata from a public data portal, the analysis provided an empirical diagnosis of how Open Data in the S&T domain embodies core Open Science principles—particularly transparency, reproducibility, and collaboration—as articulated through the FAIR framework (Wilkinson et al., 2016 ). The findings demonstrate that public S&T data form a multi-layered semantic structure encompassing the entire R&D lifecycle, rather than a fragmented or purely administrative collection of records. The identification of four core semantic axes—Research Outcomes, Research Infrastructure, Industry Support, and Raw Data—reveals that contemporary data openness extends beyond the disclosure of final outputs toward the inclusion of intermediate products and input resources generated throughout research processes. This structural configuration indicates a maturation of data-opening strategies toward ecosystem-level openness, rather than output-centric disclosure, aligning with the view that open data ecosystems must evolve to support diverse stakeholder interactions (Janssen et al., 2012 ; Zuiderwijk et al., 2014 ). Importantly, the study provides empirical evidence that public S&T data are semantically aligned with the principle of reproducibility. The presence of topics containing contextual information on experimental environments, equipment specifications, and process conditions suggests a transition from result-oriented data sharing to context-aware data provision. Such semantic characteristics enhance the potential for third-party verification, replication, and secondary analysis, thereby supporting cumulative knowledge production and the long-term sustainability of Open Science (Munafò et al., 2017 ; Murray-Rust, 2008 ). This finding empirically validates the theoretical argument that data must be accompanied by provenance information to serve as epistemic resources (Leonelli, 2016 ). At the same time, this study highlights a critical limitation in current open data practices: the persistence of structural metadata deficiencies. While descriptive metadata is often sufficiently detailed, inconsistencies in structural metadata—particularly non-standardized column names and schema designs embedded in raw datasets—significantly undermine interoperability and computational reuse. This finding underscores that semantic richness alone is insufficient to realize the full benefits of Open Science if data operability remains constrained (Kubler et al., 2018 ). It reinforces the notion that without machine-actionable standards, data remains "technically open" but "practically unusable" for large-scale integration (Mons et al., 2017 ). By shifting the analytical focus from metadata description to metadata operability, this study advances scientometric research on Open Data quality. The proposed semantic diagnostic framework illustrates how latent thematic structures and structural weaknesses can be jointly identified, offering a more nuanced understanding of data readiness for reuse than indicator-based assessments alone. In doing so, the study contributes methodologically by demonstrating the value of computational semantic analysis as a tool for evaluating the conceptual and infrastructural dimensions of Open Science (Grootendorst, 2022 ; Vicente-Saez & Martinez-Fuentes, 2018 ). From a policy and governance perspective, the results suggest that future Open Science initiatives should prioritize qualitative refinement over further quantitative expansion. Specifically, policy interventions should mandate standardized, machine-readable structural metadata alongside descriptive documentation to reduce data linkage costs, enhance reproducibility, and facilitate cross-domain and cross-sectoral data integration. Strengthening metadata operability is therefore not merely a technical improvement, but a prerequisite for translating openness into sustained scientific innovation and socio-economic value (OECD, 2015 ). Taken together, this study reinforces the view that the effectiveness of Open Science depends not only on the volume or visibility of open data, but on its capacity to be semantically interpretable and structurally interoperable. By empirically diagnosing both dimensions, the study offers a foundation for future research and policy efforts aimed at building robust, reusable, and scalable Open Science infrastructures. Declarations Author Contribution J.J. conceived the study, designed the methodology, collected and analyzed the data, and wrote the manuscript. The author read and approved the final manuscript. Data Availability Publicly available in a repository: The datasets analyzed during the current study are available in the Public Data Portal (Ministry of the Interior and Safety, Republic of Korea) at https://www.data.go.kr. References Aria, M., Misuraca, M., & Spano, M. (2020). Mapping the evolution of social research and data science on 30 years of social indicators research. Social indicators research , 149 (3), 803–831. Bertot, J. C., Jaeger, P. T., & Grimes, J. M. (2010). Using ICTs to create a culture of transparency: E-government and social media as openness and anti-corruption tools for societies. Government Information Quarterly , 27(3), 264–271. https://doi.org/10.1016/j.giq.2010.03.001 Blei, D. M., & Lafferty, J. D. (2006). Dynamic topic models. In Proceedings of the 23rd International Conference on Machine Learning (pp. 113–120). ACM. https://doi.org/10.1145/1143844.1143859 Blei, D. M., Ng, A. Y., & Jordan, M. I. (2003). Latent Dirichlet Allocation. Journal of Machine Learning Research , 3(Jan), 993–1022. https://doi.org/10.5555/944919.944937 Borgman, C. L. (2012). The conundrum of sharing research data. Journal of the American Society for Information Science and Technology , 63(6), 1059–1078. https://doi.org/10.1002/asi.22634 Borgman, C. L. (2015). Big Data, Little Data, No Data: Scholarship in the Networked World . MIT Press. Borgman, C. L. (2019a). Open data, grey data, and stewardship: Universities at the privacy frontier. Berkeley Technology Law Journal , 33(2), 365–412. Borgman, C. L. (2019b). The lives and after lives of data. Harvard Data Science Review , 1(1). https://doi.org/10.1162/99608f92.9a36bdb6 Börner, K., Chen, C., & Boyack, K. W. (2003). Visualizing knowledge domains. Annual Review of Information Science and Technology , 37(1), 179–255. Campello, R. J. G. B., Moulavi, D., & Sander, J. (2013). Density-based clustering based on hierarchical density estimates. In Pacific-Asia Conference on Knowledge Discovery and Data Mining (pp. 160–172). Springer. https://doi.org/10.1007/978-3-642-37456-2_14 Carbonell, J., & Goldstein, J. (1998). The use of MMR, diversity-based reranking for reordering documents and producing summaries. In Proceedings of the 21st ACM SIGIR Conference (pp. 335–336). ACM. https://doi.org/10.1145/290941.291025 Cho, J. (2023). Comparative analysis of open government data topics and usability. Quality & Quantity , 57 (6), 5655–5671. Dai, S. C. (2025). Conceptualizing Digital Democracy: A Policy-Oriented Approach Through Computational Text Analysis. Policy & Internet , 17 (4), e70022. Edwards, P. N., Mayernik, M. S., Batcheller, A. L., Bowker, G. C., & Borgman, C. L. (2011). Science friction: Data, metadata, and collaboration. Social Studies of Science , 41(5), 667–690. https://doi.org/10.1177/0306312713491721 Fanelli, D., et al. (2018). Is science really facing a reproducibility crisis? Proceedings of the National Academy of Sciences , 115(11), 2628–2631. Fecher, B., & Friesike, S. (2014). Open Science: One Term, Five Schools of Thought. In S. Bartling & S. Friesike (Eds.), Opening Science (pp. 17–47). Springer. https://doi.org/10.1007/978-3-319-00026-8_2 Goodman, S. N., Fanelli, D., & Ioannidis, J. P. A. (2016). What does research reproducibility mean? Science Translational Medicine , 8(341), 341ps12. https://doi.org/10.1126/scitranslmed.aaf5027 Grootendorst, M. (2022). BERTopic: Neural topic modeling with a class-based TF-IDF procedure. arXiv preprint arXiv:2203.05794 . https://doi.org/10.48550/arXiv.2203.05794 Gokhberg, L., Meissner, D., & Kuzminov, I. (2023). What semantic analysis can tell us about long term trends in the global STI policy agenda. The Journal of Technology Transfer , 48 (6), 2249–2277. Hecker, S., Haklay, M., Bowser, A., Makuch, Z., Vogel, J., & Bonn, A. (2018). Citizen Science: Innovation in Open Science, Society and Policy . UCL Press. https://doi.org/10.14324/111.9781787352339 Jacobsen, A., de Miranda Azevedo, R., Juty, N., Batista, D., Coles, S., Cornet, R., … Mons, B. (2020). FAIR principles: Interpretations and implementation considerations. Data Intelligence , 2(1–2), 10–29. https://doi.org/10.1162/dint_r_00024 Janssen, M., Charalabidis, Y., & Zuiderwijk, A. (2012). Benefits, adoption barriers and myths of open data and open government. Information Systems Management , 29(4), 258–268. https://doi.org/10.1080/10580530.2012.716740 Kitchin, R. (2014). The Data Revolution: Big Data, Open Data, Data Infrastructures and Their Consequences . Sage. Kubler, S., Robert, J., Neumaier, S., Umbrich, J., & Le Traon, Y. (2018). Open data portal quality comparison using AHP. Government Information Quarterly , 35(2), 13–29. https://doi.org/10.1016/j.giq.2018.02.001 Kučera, J., Chlapek, D., & Nečaský, M. (2013). Quality assessment of open government data. In International Conference on Electronic Government (pp. 351–359). Springer. https://doi.org/10.1007/978-3-642-40160-2_28 Lee, J. (2020). Demand forecasting of public data using keyword network analysis. Informatization Policy , 27(4), 24–46. https://doi.org/10.22693/NIAIP.2020.27.4.024 Leonelli, S. (2016). Data-Centric Biology: A Philosophical Study . University of Chicago Press. McInnes, L., Healy, J., & Melville, J. (2018). UMAP: Uniform Manifold Approximation and Projection for dimension reduction. arXiv preprint arXiv:1802.03426 . https://doi.org/10.48550/arXiv.1802.03426 Mons, B., Neylon, C., Velterop, J., Dumontier, M., da Silva Santos, L. O. B., & Wilkinson, M. D. (2017). Cloudy, increasingly FAIR; revisiting the FAIR data guiding principles for the European Open Science Cloud. Information Services & Use , 37(1), 49–56. https://doi.org/10.3233/ISU-170824 Munafò, M. R., Nosek, B. A., Bishop, D. V., Button, K. S., Chambers, C. D., Du Sert, N. P., … Ioannidis, J. P. (2017). A Manifesto for Reproducible Science. Nature Human Behaviour , 1(1), 0021. https://doi.org/10.1038/s41562-016-0021 Murray-Rust, P. (2008). Open Data in Science. Serials Review , 34(1), 52–64. https://doi.org/10.1016/j.serrev.2008.01.001 OECD. (2015). Data-Driven Innovation: Big Data for Growth and Well-Being . OECD Publishing. https://doi.org/10.1787/9789264229358-en OECD. (2021). Enhancing Access to and Sharing of Data: Reconciling Risks and Benefits for Data Re-use across Societies . OECD Publishing. https://doi.org/10.1787/276aaca8-en Pasquetto, I. V., Borgman, C. L., & Wofford, M. F. (2019). Uses and reuses of scientific data: The data creators' advantage. Proceedings of the National Academy of Sciences , 116(10), 4108–4117. https://doi.org/10.1073/pnas.1804247116 Piwowar, H. A., & Vision, T. J. (2013). Data reuse and the open data citation advantage. PeerJ , 1, e175. https://doi.org/10.7717/peerj.175 Piwowar, H. A., Priem, J., Larivière, V., Alperin, J. P., Matthias, L., Norlander, B., … Haustein, S. (2018). The state of OA: A large-scale analysis of the prevalence and impact of Open Access articles. PeerJ , 6, e4375. https://doi.org/10.7717/peerj.4375 Qiang, J., Qian, Z., Li, Y., Yuan, Y., & Wu, X. (2020). Short text topic modeling techniques, applications, and performance: A survey. IEEE Transactions on Knowledge and Data Engineering , 34(3), 1427–1445. https://doi.org/10.1109/TKDE.2020.2999485 Reimers, N., & Gurevych, I. (2019). Sentence-BERT: Sentence embeddings using Siamese BERT-networks. In Proceedings of EMNLP-IJCNLP 2019 (pp. 3982–3992). ACL. https://doi.org/10.18653/v1/D19-1410 Ruijer, E., & Meijer, A. (2020). Open government data as an innovation process: Lessons from a living lab experiment. Public Performance & Management Review , 43 (3), 613–635. Stodden, V., McNutt, M., Bailey, D. H., Deelman, E., Gil, Y., Hanson, B., … Taufer, M. (2016). Enhancing reproducibility for computational methods. Science , 354(6317), 1240–1241. https://doi.org/10.1126/science.aah6168 Tenopir, C., Dalton, E. D., Allard, S., Frame, M., Pjesivac, I., Birch, B., … Klentzens, K. (2011). Data sharing by scientists: practices and perceptions. PLoS One , 6(6), e21101. https://doi.org/10.1371/journal.pone.0021101 Thibault, R. T., Amaral, O. B., Argolo, F., Bandrowski, A. E., Davidson, A. R., & Drude, N. I. (2023). Open Science 2.0: Towards a truly collaborative research ecosystem. PLoS Biology , 21 (10), e3002362. https://doi.org/10.1371/journal.pbio.3002362 Ubaldi, B. (2013). Open government data: Towards empirical analysis of open government data initiatives. OECD Working Papers on Public Governance , (22). OECD Publishing. https://doi.org/10.1787/5k46bj4f03s7-en Vicente-Saez, R., & Martinez-Fuentes, C. (2018). Open Science now: A systematic literature review. Journal of Business Research , 88, 428–436. https://doi.org/10.1016/j.jbusres.2017.12.043 Walker, R. M., Chandra, Y., Zhang, J., & van Witteloostuijn, A. (2019). Topic modeling the research-practice gap in public administration. Public administration review , 79 (6), 931–937. Waltman, L. (2016). A review of the literature on citation impact indicators. Journal of Informetrics , 10(2), 365–391. Wilkinson, M. D., Dumontier, M., Aalbersberg, I. J., Appleton, G., Axton, M., Baak, A., … Mons, B. (2016). The FAIR Guiding Principles for scientific data management and stewardship. Scientific Data , 3, 160018. https://doi.org/10.1038/sdata.2016.18 Zuiderwijk, A., & Janssen, M. (2014). Open data policies, their implementation and impact. Government Information Quarterly , 31(1), 17–29. https://doi.org/10.1016/j.giq.2013.04.003 Zuiderwijk, A., Janssen, M., & Davis, C. (2014). Innovation with open data: Essential elements of open data ecosystems. Information Polity , 19(1, 2), 17–33. https://doi.org/10.3233/IP-140329 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8440351","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":581387024,"identity":"5220c567-baa4-4c4e-bc69-933175b21cf4","order_by":0,"name":"Junyoung Jeong","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA0ElEQVRIiWNgGAWjYBACxgYgIQFisTcwgzkMB4jWwnOASC0IIJFApBbmBh4DBosKmzz5yOePDWe2Mcjx3Ugg5DCgFokzacWGt3OMEze2MRhLEqVFsu1w4sbZOcwHH7YxJG4gTss/oJaZxx+DtNQTqaXhcOJ8CQawwxIMCGppZis4IHEsLXEDT46x4YxzEoYzzzzAr8WwvXnjY4kam8T57ccfS/aU2cjzHSdgi2Ezh8FhUFQaHADzJfArBwF5BvYHjB9AjAbCikfBKBgFo2CEAgD0Q0bEkmr3zwAAAABJRU5ErkJggg==","orcid":"","institution":"National Information Society Agency","correspondingAuthor":true,"prefix":"","firstName":"Junyoung","middleName":"","lastName":"Jeong","suffix":""}],"badges":[],"createdAt":"2025-12-24 07:53:50","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8440351/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8440351/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":101751529,"identity":"e2e4b9e6-c06a-43b1-b389-12e7904af596","added_by":"auto","created_at":"2026-02-03 10:21:05","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":94557,"visible":true,"origin":"","legend":"\u003cp\u003eMechanism of the automated machine learning used in this study\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8440351/v1/26ed7cf22f186c7fb3902b04.png"},{"id":101440846,"identity":"cd25b9c4-1751-499a-8f09-faea65eacd48","added_by":"auto","created_at":"2026-01-29 16:57:23","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":101686,"visible":true,"origin":"","legend":"\u003cp\u003eResearch procedure framework: From data collection to semantic analysis\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8440351/v1/5fcab1f55fed2d80bfe22b79.png"},{"id":101440847,"identity":"89b9569c-1a85-4724-ab52-d4d5be410836","added_by":"auto","created_at":"2026-01-29 16:57:23","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":105633,"visible":true,"origin":"","legend":"\u003cp\u003eHierarchical semantic similarity among major topic clusters in public S\u0026amp;T data\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8440351/v1/d832065d4c0fd83301268004.png"},{"id":101440845,"identity":"c40a764d-c122-4526-bba1-78ba70870a86","added_by":"auto","created_at":"2026-01-29 16:57:23","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":195008,"visible":true,"origin":"","legend":"\u003cp\u003eSemantic topology and spatial distribution of topic clusters in public S\u0026amp;T data\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8440351/v1/a8fe2207a39551c72427521a.png"},{"id":101754863,"identity":"06b931ba-b2ba-45ac-a388-1d61e33d1907","added_by":"auto","created_at":"2026-02-03 10:47:23","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1403829,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8440351/v1/ac2d5746-99c4-4416-938a-11b87f4cba4d.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"From Open Data to Open Science?: A Semantic Diagnosis of Public Science and Technology Data","fulltext":[{"header":"1. INTRODUCTION","content":"\u003cp\u003eWith the acceleration of the Fourth Industrial Revolution, data has become a strategic asset underpinning scientific progress, economic competitiveness, and evidence-based policymaking. Frequently described as the \u0026ldquo;oil of the 21st century,\u0026rdquo; data\u0026mdash;particularly that generated or held by governments\u0026mdash;has been increasingly released as Open Data to enhance transparency, accountability, and innovation (Janssen et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Bertot et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). International organizations such as the OECD have consistently emphasized that open government data constitutes a foundational infrastructure for data-driven innovation and inclusive growth (OECD, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; OECD, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2021\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eSouth Korea represents a leading case in the global Open Data landscape. Following the enactment of the \u003cem\u003eAct on Promotion of the Provision and Use of Public Data\u003c/em\u003e, the Korean government rapidly expanded the scale of public data disclosure through its national Open Data Portal (data.go.kr). While this quantitative expansion successfully lowered access barriers and stimulated early-stage data utilization, recent scholarship increasingly questions whether mere availability translates into meaningful reuse and value creation (Zuiderwijk \u0026amp; Janssen, \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Zuiderwijk et al., \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). As artificial intelligence (AI), machine learning, and large-scale data integration become central to scientific inquiry, qualitative dimensions such as usability, interoperability, and structural consistency have emerged as critical bottlenecks (Borgman, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Edwards et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2011\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThese challenges are particularly salient in the science and technology (S\u0026amp;T) domain, where public data disclosure is closely intertwined with the broader paradigm of Open Science. Open Science extends beyond open access to publications, advocating transparency across the entire research lifecycle, including data, methods, software, and workflows (Fecher \u0026amp; Friesike, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Vicente-Saez \u0026amp; Martinez-Fuentes, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). From this perspective, publicly funded research data is not merely an administrative byproduct but a core scientific infrastructure enabling reproducibility, cumulative knowledge production, and interdisciplinary convergence (Leonelli, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Stodden et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eWithin the Open Science discourse, the FAIR Guiding Principles\u0026mdash;Findable, Accessible, Interoperable, and Reusable\u0026mdash;have become the dominant evaluative framework for assessing data stewardship quality (Wilkinson et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). Subsequent studies have further emphasized that interoperability and reusability are decisive conditions for machine-actionable science, particularly in AI-driven research environments (Jacobsen et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Mons et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Data that lacks standardized schemas, controlled vocabularies, or consistent column-level semantics remains difficult to integrate, regardless of its formal openness, effectively functioning as a \u0026ldquo;data dump\u0026rdquo; rather than a reusable scientific asset (Murray-Rust, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2008\u003c/span\u003e; Borgman, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2019a\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eDespite the conceptual centrality of FAIR principles, empirical research on public data\u0026mdash;especially within the S\u0026amp;T sector\u0026mdash;has remained skewed toward policy evaluation, portal-level performance metrics, or user satisfaction surveys (Ubaldi, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). While recent studies have applied text mining and topic modeling to analyze descriptive metadata such as titles and abstracts (Lee, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), far less attention has been paid to \u003cem\u003estructural metadata\u003c/em\u003e, including column names, variable definitions, and implicit semantic hierarchies. This omission is nontrivial: in data-intensive scientific domains, inconsistencies at the column level directly undermine interoperability, reproducibility, and cross-domain integration (Kučera et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Munaf\u0026ograve; et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Goodman et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eTo address this gap, the present study adopts a scientometric and computational semantic perspective to examine how public S\u0026amp;T data is structurally positioned within the discourse of Open Science. Rather than treating Open Data as a static policy output, we conceptualize it as a semantic infrastructure whose internal organization reflects\u0026mdash;and potentially constrains\u0026mdash;the realization of Open Science principles (Blei \u0026amp; Lafferty, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2006\u003c/span\u003e; Hecker et al., \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Methodologically, we employ BERTopic, a state-of-the-art neural topic modeling approach that leverages contextualized embeddings to overcome the sparsity and polysemy limitations of traditional Latent Dirichlet Allocation (Blei et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2003\u003c/span\u003e; Qiang et al., \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Grootendorst, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eBy extending the analytical scope beyond descriptive text to include key column names and hierarchical topic structures, this study empirically maps the semantic topology of public S\u0026amp;T data. Through dimensionality reduction (UMAP), density-based clustering (HDBSCAN), and diversity-aware keyword extraction (MMR), we investigate whether and how current public data practices support the FAIR principles\u0026mdash;particularly interoperability and reusability\u0026mdash;at a structural level (Reimers \u0026amp; Gurevych, \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; McInnes et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Campello et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Carbonell \u0026amp; Goldstein, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e1998\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eUltimately, this study seeks to bridge the gap between Open Data policy ambitions and their practical realization within Open Science. By revealing the latent semantic structure and positional centrality of S\u0026amp;T data topics, we aim to answer a fundamental question confronting contemporary data governance: \u003cb\u003eDoes Open Data, as currently implemented, genuinely function as an enabling infrastructure for Open Science?\u003c/b\u003e\u003c/p\u003e \u003cp\u003eAccordingly, this study addresses the following research questions:\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003eRQ1.\u003c/b\u003e What semantic roles does Open Data play within the discourse of Open Science in the science and technology sector?\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003eRQ2.\u003c/b\u003e To what extent does Open Data occupy a central or peripheral position in the semantic structure of Open Science?\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003eRQ3.\u003c/b\u003e How does the semantic positioning of Open Data relate to key Open Science principles\u0026mdash;particularly the FAIR principles\u0026mdash;across domains and over time?\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e"},{"header":"2. Theoretical Background","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1. Open Data, Metadata, and the Problem of Usability\u003c/h2\u003e \u003cp\u003eOpen Data is broadly defined as data that can be freely accessed, used, modified, and redistributed by any actor without technical or legal restrictions. In the context of public governance, Open Data has been widely conceptualized as a key infrastructural resource for enhancing transparency, accountability, and innovation across sectors (Janssen et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Bertot et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). International policy organizations, particularly the OECD, have framed Open Data as a foundational component of the digital economy, emphasizing its potential to generate social value, stimulate economic growth, and enable data-driven policy-making (Ubaldi, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; OECD, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; OECD, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2021\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eFollowing the enactment of the Act on Promotion of the Provision and Use of Public Data, South Korea has pursued comprehensive, government-wide open data initiatives, achieving world-leading performance in quantitative indicators such as the number of released datasets and international open data rankings. However, as open data policies have matured, scholarly attention has increasingly shifted from questions of availability toward deeper concerns regarding usability, interpretability, and data quality (Zuiderwijk \u0026amp; Janssen, \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Zuiderwijk et al., \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2014\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eA growing body of literature suggests that the mere release of datasets does not guarantee meaningful reuse. Empirical studies have shown that users frequently encounter difficulties in understanding, integrating, and repurposing open data due to insufficient contextualization and inconsistent data structures (Tenopir et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Kubler et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). This recognition has repositioned metadata as a central determinant of open data usability and value creation (Kučera et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Borgman, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2015\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eMetadata\u0026mdash;commonly defined as \u0026ldquo;data about data\u0026rdquo;\u0026mdash;provides critical information regarding the content, structure, provenance, and conditions of use of datasets (Borgman, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Mons et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Prior research has conceptualized metadata as a multi-layered construct encompassing descriptive, structural, and administrative dimensions (Kučera et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Kubler et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Descriptive metadata, such as titles, abstracts, and keywords, primarily facilitates dataset discovery and thematic identification. Structural metadata, including schemas, variable definitions, data types, and column names, governs the technical feasibility of machine processing, interoperability, and large-scale integration (Wilkinson et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Jacobsen et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eDespite this distinction, existing domestic and international studies have predominantly focused on evaluating the completeness and accuracy of descriptive metadata. For example, Lee (\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) applied text mining techniques to metadata descriptions to identify thematic biases in public data portals, while Kučera et al. (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2013\u003c/span\u003e) proposed quantitative indicators to assess missing values and syntactic correctness within metadata fields. In contrast, systematic analyses of structural metadata\u0026mdash;particularly the semantic consistency and standardization of column names embedded within raw datasets\u0026mdash;remain relatively underdeveloped.\u003c/p\u003e \u003cp\u003eThe absence of standardized structural metadata substantially increases preprocessing costs, limits cross-domain integration, and undermines the cumulative reuse of data (Zuiderwijk et al., \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Kubler et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). From a scientometric perspective, such structural fragmentation constrains the formation of scalable knowledge infrastructures and diminishes the long-term scientific and economic value of Open Data (Edwards et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Borgman, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2019a\u003c/span\u003e). Addressing this gap requires analytical approaches that move beyond surface-level metadata descriptions to diagnose the deeper semantic and structural organization of open data ecosystems.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2. Open Science and FAIR-Oriented Data Infrastructures\u003c/h2\u003e \u003cp\u003eWithin the science and technology domain, data disclosure is embedded in the broader paradigm of Open Science. Open Science represents a systemic transformation of the research lifecycle, extending beyond Open Access publications to encompass the sharing of raw data, methodologies, software, and intermediate research outputs (Fecher \u0026amp; Friesike, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Vicente-Saez \u0026amp; Martinez-Fuentes, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). This movement reflects both a normative commitment to science as a public good and a pragmatic response to increasing concerns regarding research transparency and reproducibility (Munaf\u0026ograve; et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Stodden et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eResearch data occupy a central position in Open Science infrastructures, as they enable verification, replication, and cumulative knowledge production (Leonelli, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Goodman et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). Empirical studies have demonstrated that openly shared data are associated with higher citation rates and increased scientific impact, reinforcing the instrumental value of Open Data within scientific ecosystems (Piwowar \u0026amp; Vision, \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Piwowar et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2018\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe FAIR Guiding Principles\u0026mdash;Findable, Accessible, Interoperable, and Reusable\u0026mdash;have emerged as the de facto international standard for scientific data stewardship (Wilkinson et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). Unlike earlier openness frameworks, FAIR explicitly emphasizes machine-actionability and interoperability as prerequisites for scalable data reuse (Mons et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Jacobsen et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Among the four principles, interoperability is particularly critical, as it requires data to be described using standardized vocabularies, shared schemas, and consistent structural metadata.\u003c/p\u003e \u003cp\u003eWhen datasets generated by different research groups employ harmonized column names, variable definitions, and measurement units, integrated analysis can proceed with minimal friction. Conversely, datasets lacking such structural alignment remain isolated \u0026ldquo;data silos,\u0026rdquo; severely limiting their contribution to reproducible and data-intensive science (Leonelli, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Murray-Rust, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2008\u003c/span\u003e). Studies on reproducible research further indicate that access to raw data alone is insufficient if the structural and contextual information necessary to reconstruct analytical processes is absent (Munaf\u0026ograve; et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Stodden et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eDespite the centrality of Open Data in Open Science discourse, existing studies have largely focused on policy declarations, compliance indicators, or normative frameworks. Relatively little attention has been paid to how Open Data is semantically positioned within large-scale data infrastructures and whether it functions as a structurally enabling component of Open Science in practice. This gap motivates the need for empirical analyses that examine how Open Data is organized, clustered, and conceptually related to core Open Science principles across domains.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3. Computational Topic Modeling and Semantic Analysis Using BERTopic\u003c/h2\u003e \u003cp\u003eTo investigate the latent semantic organization of large-scale textual data, researchers have employed a range of topic modeling techniques. Latent Dirichlet Allocation (LDA) has long served as the dominant approach, modeling documents as probabilistic mixtures of topics based on word co-occurrence patterns (Blei et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2003\u003c/span\u003e). However, LDA\u0026rsquo;s \u0026ldquo;bag-of-words\u0026rdquo; assumption prevents it from capturing contextual meaning and word order, rendering it vulnerable to polysemy and semantic ambiguity (Blei \u0026amp; Lafferty, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2006\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThese limitations are particularly pronounced in the analysis of short texts, such as dataset titles, keywords, and metadata descriptions, where sparsity often leads to incoherent or unstable topics (Qiang et al., \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). To address these challenges, recent studies have increasingly adopted neural embedding-based approaches that preserve contextual semantics.\u003c/p\u003e \u003cp\u003eBERTopic represents a state-of-the-art topic modeling framework that integrates transformer-based language models with density-based clustering techniques (Grootendorst, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). By embedding documents using Sentence-BERT (Reimers \u0026amp; Gurevych, \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), BERTopic captures semantic similarity at the sentence level. Dimensionality reduction is performed using UMAP (McInnes et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2018\u003c/span\u003e), followed by clustering with HDBSCAN, which enables the identification of hierarchical topic structures without requiring a predefined number of topics (Campello et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). Representative keywords are extracted using class-based TF-IDF (c-TF-IDF), with diversity further enhanced through Maximal Marginal Relevance (Carbonell \u0026amp; Goldstein, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e1998\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eBecause BERTopic clusters documents based on semantic proximity rather than lexical frequency, it is particularly well-suited for analyzing heterogeneous metadata environments where descriptive and structural elements are interwoven. Moreover, its capacity to visualize hierarchical and spatial topic relationships aligns with the objectives of scientometric research that seeks to uncover both micro-level thematic patterns and macro-level structural configurations. Accordingly, this study employs BERTopic to empirically diagnose the semantic topology of public science and technology Open Data and to assess its structural alignment with the principles of Open Science.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. MATERIALS AND METHODS","content":"\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e3.1. Data Collection and Preprocessing\u003c/h2\u003e \u003cp\u003eTo empirically analyze the semantic structure and standardization status of Open Data in the domestic science and technology sector, this study selected the \"Public Data Registration Management System,\" operated by the Ministry of the Interior and Safety, as the primary source for data collection. This system serves as a government-wide platform that registers and manages metadata for data held by central ministries, local governments, and public institutions, providing an optimal sample for identifying the status of national data initiatives. Data collection was conducted as of December 23, 2024, targeting all file data registered under the \"Science and Technology\" category within the system's classification scheme. Although a total of 4,562 data entries were identified, the raw data contained significant noise and a high volume of entries with low metadata fidelity, rendering them unsuitable for topic modeling.\u003c/p\u003e \u003cp\u003eConsequently, a stepwise data cleaning process was implemented to ensure analysis reliability and optimize the performance of the BERTopic model. First, De-identification of Institution Names was performed. Public data titles frequently include the names of providing institutions, such as \"[Ministry of Environment],\" \"(Foundation) Research Institute,\" or \"OO Center.\" Since these names are irrelevant to the intrinsic subject matter and can induce bias where data clusters around \"specific institutions\" rather than topics, we utilized Regular Expressions to systematically remove text within square brackets [] and parentheses () along with the enclosed institution names, retaining only the pure keywords.\u003c/p\u003e \u003cp\u003eSecond, Normalization and Concatenation were conducted. After removing unnecessary special characters and excessive spacing, three metadata fields\u0026mdash;1) Title, 2) Description, and 3) Keywords \u0026amp; Column Names\u0026mdash;were concatenated into a single document. This step was taken to mitigate the sparsity problem inherent in short texts and to enable the BERTopic model to learn the semantic context more robustly. Third, Morphological Analysis was performed. Given the agglutinative nature of the Korean language, tokenization based simply on spacing has limitations in capturing meaning. Therefore, we utilized the Okt (Open Korean Text) morphological analyzer to extract nouns, which are the core parts of speech constituting the meaning of sentences.\u003c/p\u003e \u003cp\u003eThrough this rigorous preprocessing pipeline, 1,070 data entries were finally confirmed as the valid analysis corpus. This corresponds to approximately 23.5% of the initially collected data, a selection made to enhance the precision of the analysis results by utilizing only high-quality text data from which noise had been eliminated.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e3.2. BERTopic Methodology\u003c/h2\u003e \u003cp\u003eThis study utilized the BERTopic algorithm to extract latent topics and capture their nuanced context from the collected unstructured text data. BERTopic is a sophisticated model proposed to overcome the \"Bag-of-Words\" limitations inherent in traditional probabilistic models like Latent Dirichlet Allocation (LDA), which typically ignores word order and broader semantic context. By combining pre-trained Transformer-based embeddings with a class-based TF-IDF (c-TF-IDF) procedure, the model maximizes the semantic consistency and coherence of the derived topics.\u003c/p\u003e \u003cp\u003eRegarding the specific analysis procedure, we first applied the SBERT (Sentence-BERT) framework to transform each document into a dense, high-dimensional vector space. To accurately capture the intricate contextual nuances of Korean text, we employed the paraphrase-multilingual-MiniLM-L12-v2 model. This specific model is specialized for multilingual processing and is designed to position semantically similar sentences closely within the vector space, thereby mitigating the data sparsity issues common in short-text metadata.\u003c/p\u003e \u003cp\u003eSince the generated embedding vectors possess high dimensionality (over 384 dimensions), the UMAP (Uniform Manifold Approximation and Projection) algorithm was performed to prevent the \"curse of dimensionality\" that often occurs during density-based clustering. UMAP is a non-linear dimensionality reduction technique that effectively reduces dimensions while preserving both the local and global topological structure of the data. In this study, to preserve semantic similarity for subsequent clustering, dimensions were reduced to 5 components using cosine distance, with the number of neighbors (n_neighbors) set to 15 to balance the focus between local and global structures.\u003c/p\u003e \u003cp\u003eFor clustering the dimensionally reduced vectors, HDBSCAN (Hierarchical Density-Based Spatial Clustering of Applications with Noise) was applied. HDBSCAN offers a significant advantage over partition-based methods (like K-Means) as it does not require a prior specification of the number of clusters and enhances topic clarity by treating low-density outliers as noise. To prevent the generation of excessively fragmented or microscopic topics, the minimum cluster size (min_cluster_size) was set to 10.\u003c/p\u003e \u003cp\u003eFinally, the keywords representing each clustered topic were extracted using the c-TF-IDF method. This method calculates word importance by considering clusters, rather than individual documents, as the unit of analysis.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e3.3. Research Procedure\u003c/h2\u003e \u003cp\u003eThis study established and executed a systematic three-stage framework, as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, to identify the semantic structure of public data in the science and technology sector and empirically analyze how it connects to the actual Open Science ecosystem.\u003c/p\u003e \u003cp\u003eThe first stage, Data Collection, involves the selection and acquisition of raw datasets in the science and technology field, which serve as the subject of analysis. The data collected in this process goes beyond physical aggregation; it is checked for standardization status and refined into an analyzable form to serve as the foundational material for subsequent analysis.\u003c/p\u003e \u003cp\u003eThe second stage is Topic Modeling, which begins with a preprocessing step to improve the quality of the collected unstructured text data. After stop-word removal and morphological analysis, the text data is input into the BERTopic algorithm, the core methodology of this study. Through a sequence of embedding, dimensionality reduction, and clustering, latent topics within the text are derived, thereby concretizing the semantic structure of the science and technology data.\u003c/p\u003e \u003cp\u003e The final third stage is Topic Analysis. For each topic generated through modeling, Topic Labeling is performed by reviewing key keywords and representative documents to assign subject names. This is followed by a Topic Classification process that categorizes themes of a similar nature. Finally, the study derives its conclusions by performing an Evaluation of Linkage to Open Science, which analyzes how well these structured topics align with the core values of Open Science\u0026mdash;openness, sharing, and collaboration\u0026mdash;and determines whether current data disclosure leads to the revitalization of Open Science.\u003c/p\u003e \u003cp\u003e \u003cb\u003eFigure\u0026nbsp;2.\u003c/b\u003e Research procedure framework: From data collection to semantic analysis\u003c/p\u003e \u003c/div\u003e"},{"header":"4. RESULTS","content":"\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e4.1. Topic Modeling Analysis\u003c/h2\u003e \u003cp\u003eThis study applied BERTopic modeling (Grootendorst, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) to public science and technology (S\u0026amp;T) datasets to uncover latent semantic structures embedded in large-scale metadata. Compared to probabilistic topic modeling approaches such as Latent Dirichlet Allocation (LDA), which are known to suffer from sparsity and semantic dilution when applied to short and heterogeneous metadata texts (Blei et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2003\u003c/span\u003e; Qiang et al., \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), BERTopic leverages contextual sentence embeddings to capture semantic similarity at a higher conceptual resolution. Recent scientometric studies have emphasized that embedding-based topic modeling is particularly suitable for policy-oriented and infrastructure-related corpora, where terminological variation is high and semantic nuance is critical (Dai, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Walker et al., \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe final model achieved a silhouette coefficient of 0.681, indicating strong topic separability and internal coherence. This level of clustering performance is comparable to or exceeds benchmarks reported in recent large-scale semantic analyses of research policy documents and public data repositories (Gokhberg et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Aria et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). After excluding noise documents identified through density-based clustering, a total of 30 coherent topics were extracted, providing a granular semantic representation of the national public S\u0026amp;T data landscape.\u003c/p\u003e \u003cp\u003eA closer examination of topic keywords reveals that Topic 0 (Intellectual Property \u0026amp; Patent Status) occupies a structurally prominent position, characterized by high-frequency terms such as \u003cem\u003eregistration number\u003c/em\u003e, \u003cem\u003eapplication number\u003c/em\u003e, and \u003cem\u003epublication number\u003c/em\u003e. This dominance suggests that a substantial portion of public S\u0026amp;T data is oriented toward the legal and institutional management of research outputs. Prior studies have noted that intellectual property\u0026ndash;related metadata often functions as an interface between public research systems and market-oriented innovation ecosystems, reinforcing the economic rationales underlying open data policies (Kitchin, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Ruijer \u0026amp; Meijer, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eSimilarly, topics related to academic communication\u0026mdash;such as Topic 7 (Academic Literature \u0026amp; Bibliographic Info), Topic 18 (Theses \u0026amp; Research Publications), and Topic 19 (Digital Content Metadata)\u0026mdash;are defined by descriptors including \u003cem\u003eauthor\u003c/em\u003e, \u003cem\u003epaper title\u003c/em\u003e, and \u003cem\u003eattachment link\u003c/em\u003e. These topics confirm that traditional scholarly outputs remain a foundational component of public data infrastructures, consistent with earlier findings that open bibliographic metadata serves as the backbone of knowledge dissemination and scientometric analysis (Borgman, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Waltman, \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eFrom the perspective of research reproducibility, the identification of multiple topics containing raw observational and experimental data is particularly noteworthy. Topics such as Topic 16 (Weather \u0026amp; Environment Observation), Topic 23 (Bio-health \u0026amp; Body Composition), and Topic 4 (Network Quality \u0026amp; Speed) include concrete measurement variables (e.g., \u003cem\u003ewind speed\u003c/em\u003e, \u003cem\u003ebody fat mass\u003c/em\u003e, \u003cem\u003eMbps\u003c/em\u003e), which are essential for independent verification and secondary analysis. This finding aligns with recent empirical work emphasizing that the availability of machine-readable raw data, rather than summarized results alone, is a prerequisite for cumulative and reproducible science (Stodden et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Goodman et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Fanelli et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2018\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eTopics related to research infrastructure and experimental context\u0026mdash;such as Topic 6 (Research Equipment \u0026amp; Specifications) and Topic 13 (Manufacturing Process \u0026amp; Part Data)\u0026mdash;contain detailed attributes including \u003cem\u003emodel name\u003c/em\u003e, \u003cem\u003emanufacturer\u003c/em\u003e, and \u003cem\u003eprocess code\u003c/em\u003e. These structural descriptors play a critical role in documenting the conditions under which data are generated, thereby supporting interpretability and reuse. Recent Open Science literature has increasingly highlighted the importance of such contextual metadata as a bridge between openness and actual scientific usability (Edwards et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Leonelli, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Pasquetto et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAdministrative and managerial dimensions of the national R\u0026amp;D system are also strongly represented. Topics such as Topic 3 (ICT R\u0026amp;D Projects \u0026amp; SW Development), Topic 15 (Lab \u0026amp; Project Execution Info), and Topic 21 (Institutional Research Performance Indicators) include identifiers related to project lifecycle management, performance evaluation, and institutional accountability. These findings support the view that public S\u0026amp;T data functions not only as a scientific resource but also as a governance instrument that enhances transparency and coordination within complex research systems (Bertot et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2010\u003c/span\u003e; Janssen et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Kitchin, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2014\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAt the same time, the emergence of Topic 22 and Topic 28\u0026mdash;both characterized exclusively by the absence of identifiable column information\u0026mdash;reveals persistent structural weaknesses in metadata provision. The presence of such topics corroborates prior assessments that incomplete or non-standardized structural metadata constitutes a major bottleneck for interoperability and reuse (Kubler et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Kučera et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). Recent studies further argue that these deficiencies limit the machine-actionability of open data, thereby constraining its integration into advanced analytical pipelines and AI-driven research environments (Wilkinson et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Mons et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Jacobsen et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eOverall, the 30 extracted topics can be interpreted as reflecting four interrelated pillars of the public S\u0026amp;T ecosystem: R\u0026amp;D activities, research infrastructure, industry and innovation support, and socio-cultural services. The semantic diversity and concreteness of the identified keywords suggest that current data opening practices are moving beyond symbolic disclosure toward more operational forms of openness. This structural heterogeneity mirrors contemporary conceptualizations of Open Science as a multifaceted system encompassing epistemic, institutional, and infrastructural dimensions (Fecher \u0026amp; Friesike, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Vicente-Saez \u0026amp; Martinez-Fuentes, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2018\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eList of Derived Topics and Top 5 Keywords\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTopic Label\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTop 5 Keywords\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eIntellectual Property \u0026amp; Patent Status\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRegistration number, Application number, Registration date, Title, Publication number\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLocation-based Facilities/Business Info\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePhone number, Address, Longitude, Latitude, Business name\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eForest \u0026amp; Biological Resources Info\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eKorea Forest Service, Title, Scientific name, WonPyongOh1993, Common name\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eICT R\u0026amp;D Projects \u0026amp; SW Development\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eProject number, Participating agency, Participation type, Project name, Classification code\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNetwork Quality \u0026amp; Speed\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDownload, Upload, mbps, Carrier, bit\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePostal Service \u0026amp; Reference Data\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eData reference date, October, November, December, January\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eResearch Equipment \u0026amp; Specifications\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eModel name, Sequence number, Equipment name, Manufacturer, Acquisition date\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAcademic Literature \u0026amp; Bibliographic Info\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eAuthor, Signature, Paper title, Title, Publication year\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSystem Log \u0026amp; Education History\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRegistration date/time, Usage status, Business number, Education name, Institution name\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSci-Tech Human Resource Training\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTraining project, Participation, Graduate student, Future talent, Institution name\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOnline Board Activity Status\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSerial number, Registration date, Views, Title, url\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBroadcast Ad Market Statistics\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYear, Type, Ad type, Sales, Million KRW\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTech Transfer \u0026amp; Corporate Support\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCumulative, Paid, Company, Free, By target\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 13\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eManufacturing Process \u0026amp; Part Data\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRegistration date, Substrate size, Equipment name, Process name, Process code\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 14\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAdmin Contact \u0026amp; Civil Service\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eManager, Department, Project name, Support details, Contact info\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLab \u0026amp; Project Execution Info\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLaboratory, Project classification, Lab name, Researcher name, Location\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eWeather \u0026amp; Environment Observation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eWind speed, Wind direction, edr, 50m, 76m\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eResearch Agencies \u0026amp; Duration\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eInstitution, Execution period, Data title, Data type, File name\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTheses \u0026amp; Research Publications\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTitle, Sequence number, Address, Thesis title, Language\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 19\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDigital Content Metadata\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eContent, Title, Views, Attachment link, Link\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePublic Facility Location Info\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLongitude, Latitude, Facility type, Installation place name, Address\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eInstitutional Research Performance Indicators\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eInstitution name, Type, Data name, Link, Region\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 22\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eUnidentifiable Data\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eAbsence of column info (Single keyword)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 23\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBio-health \u0026amp; Body Composition\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eWater, Extracellular water total water ratio, Item, Body fat mass, Max\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 24\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eR\u0026amp;D Workforce Demographics\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFemale, Male, Ratio, PhD, Master\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 25\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eData Lifecycle History\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDeletion status, Technology overview, Link, Future issue, Classification\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 26\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRadio Management \u0026amp; Station Permit\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePermit, Management office, Classification code, Gangneung Radio Management Office, MSIT\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 27\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBroadcast Programming \u0026amp; Channel Info\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eBroadcast field, Program name, Broadcaster name, Channel number, Nationality\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 28\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eUnidentifiable Data\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eAbsence of column info (Single keyword)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopic 29\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBroadcast Content Genre Classification\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRatio, Education, News/Current affairs, Documentary, Drama\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e4.2. Thematic Classification and Semantic Structure of Science and Technology Data\u003c/h2\u003e \u003cp\u003eThe resulting dendrogram (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e3\u003c/span\u003e) reveals that public S\u0026amp;T data do not exist as isolated informational units, but rather form an interconnected semantic ecosystem. This ecosystem exhibits a cyclic topology structured around three overarching dimensions: (1) R\u0026amp;D execution and knowledge production, (2) industrial application and innovation diffusion, and (3) social dissemination and public engagement. Such a configuration empirically supports the conceptualization of Open Science as a systemic process rather than a collection of independent openness practices (Kitchin, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Vicente-Saez \u0026amp; Martinez-Fuentes, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2018\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eFirst, the R\u0026amp;D lifecycle constitutes the semantic core of the ecosystem. Cluster 2 (R\u0026amp;D Project \u0026amp; Performance Management) aggregates administrative and operational data related to project execution, funding duration, and institutional responsibility. Its close semantic linkage with Cluster 1 (IP \u0026amp; Research Outcomes) demonstrates that research outputs\u0026mdash;such as patents, publications, and theses\u0026mdash;are structurally embedded within administrative data flows. This finding reinforces prior evidence that transparency in R\u0026amp;D management data is critical for mitigating redundant public investment and enhancing accountability in national research systems (Ruijer \u0026amp; Meijer, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Ubaldi, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). Furthermore, the adjacency of Cluster 3 (Research Infrastructure \u0026amp; Standardization) highlights the infrastructural preconditions for reproducible research. The inclusion of equipment specifications, manufacturing processes, and biological resource metadata indicates that openness is extending beyond final outputs toward the disclosure of experimental and procedural contexts, a key requirement for computational and empirical reproducibility (Stodden et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Goodman et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eSecond, the industry and innovation diffusion domain emerges as a distinct yet tightly coupled layer. Cluster 4 (Industry Support \u0026amp; Technology Transfer) captures the semantic intersection between public R\u0026amp;D and private-sector utilization, encompassing technology transfer outcomes, corporate support metrics, and data lifecycle histories. This cluster operationalizes the notion of data-driven innovation, wherein public sector information acts as a foundational input for economic value creation (OECD, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Ruijer \u0026amp; Meijer, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Cluster 7 (ICT \u0026amp; Communication Infrastructure), positioned as a supporting layer, underscores the role of network quality, radio management, and digital connectivity as enabling conditions for both industrial innovation and large-scale data reuse. Prior studies have emphasized that without interoperable and accessible digital infrastructure, the economic potential of open data remains largely unrealized (Kubler et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Zuiderwijk et al., \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2014\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThird, the presence of basic science and raw observational data signifies a qualitative maturation of public S\u0026amp;T data openness. Cluster 6 (Basic Science \u0026amp; Observational Data) includes meteorological measurements, environmental observations, and bio-health variables. Unlike aggregated statistical indicators, these datasets retain high analytical granularity, allowing downstream users to repurpose data for novel scientific inquiries. This characteristic directly corresponds to the Reusability dimension of the FAIR principles and reflects a shift from symbolic openness toward functional openness (Wilkinson et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Pasquetto et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Empirical research has shown that access to raw data substantially increases citation impact and accelerates cumulative knowledge production (Piwowar \u0026amp; Vision, \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Fanelli et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2018\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eFinally, the social diffusion and public interface layer expands the scope of S\u0026amp;T data beneficiaries beyond expert communities. Cluster 5 (Sci-Tech Human Resource Training) and Cluster 9 (Administrative \u0026amp; Civil Services) represent human capital development and public-facing administrative information, while Cluster 10 (Regional \u0026amp; Location-based Information) spatially anchors scientific resources within local contexts. The emergence of Cluster 8 (Broadcasting \u0026amp; Digital Content) further illustrates the convergence of S\u0026amp;T data with media ecosystems, transforming scientific information into consumable cultural content. This multi-layered structure substantiates arguments that Open Science increasingly functions as an infrastructure for citizen engagement, policy learning, and societal sensemaking (Hecker et al., \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Pasquetto et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn sum, the ten Macro-Topics identified through hierarchical clustering empirically demonstrate a virtuous semantic cycle in public S\u0026amp;T data. Research resources and infrastructures (Clusters 3 and 6) are mobilized through R\u0026amp;D activities (Clusters 2 and 5), generating formal outputs (Cluster 1) that diffuse into industrial (Clusters 4 and 7) and societal domains (Clusters 8\u0026ndash;10). This structural topology provides concrete evidence that semantic organization plays a pivotal role in determining whether open data can effectively evolve into Open Science.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eHierarchical Clustering Results: Macro-Topics and Constituent Micro-Topics\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCluster\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMacro-Label\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMicro-Topics (Topic ID \u0026amp; Label)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eIP \u0026amp; Research Outcomes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTopic 0 (Intellectual Property \u0026amp; Patent Status)\u003c/p\u003e \u003cp\u003eTopic 7 (Academic Literature \u0026amp; Bibliographic Info)\u003c/p\u003e \u003cp\u003eTopic 18 (Theses \u0026amp; Research Publications)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eR\u0026amp;D Project \u0026amp; Performance Management\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTopic 3 (ICT R\u0026amp;D Projects \u0026amp; SW Development)\u003c/p\u003e \u003cp\u003eTopic 15 (Lab \u0026amp; Project Execution Info)\u003c/p\u003e \u003cp\u003eTopic 17 (Research Agencies \u0026amp; Duration)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eResearch Infrastructure \u0026amp; Standardization\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTopic 6 (Research Equipment \u0026amp; Specifications)\u003c/p\u003e \u003cp\u003eTopic 13 (Manufacturing Process \u0026amp; Part Data)\u003c/p\u003e \u003cp\u003eTopic 2 (Forest \u0026amp; Biological Resources Info)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eIndustry Support \u0026amp; Tech Transfer\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTopic 11 (Broadcast Ad Market Statistics)\u003c/p\u003e \u003cp\u003eTopic 12 (Tech Transfer \u0026amp; Corporate Support)\u003c/p\u003e \u003cp\u003eTopic 25 (Data Lifecycle History)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSci-Tech Human Resource Training\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTopic 9 (Sci-Tech Human Resource Training)\u003c/p\u003e \u003cp\u003eTopic 24 (R\u0026amp;D Workforce Demographics)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBasic Science \u0026amp; Observational Data\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTopic 16 (Weather \u0026amp; Environment Observation)\u003c/p\u003e \u003cp\u003eTopic 23 (Bio-health \u0026amp; Body Composition)\u003c/p\u003e \u003cp\u003eTopic 5 (Postal Service \u0026amp; Reference Data)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eICT \u0026amp; Comm. Infrastructure\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTopic 4 (Network Quality \u0026amp; Speed)\u003c/p\u003e \u003cp\u003eTopic 26 (Radio Management \u0026amp; Station Permit)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBroadcasting \u0026amp; Digital Content\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTopic 27 (Broadcast Programming \u0026amp; Channel Info)\u003c/p\u003e \u003cp\u003eTopic 29 (Broadcast Content Genre Classification)\u003c/p\u003e \u003cp\u003eTopic 19 (Digital Content Metadata)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAdmin \u0026amp; Civil Services\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTopic 8 (System Log \u0026amp; Education History)\u003c/p\u003e \u003cp\u003eTopic 10 (Online Board Activity Status)\u003c/p\u003e \u003cp\u003eTopic 14 (Admin Contact \u0026amp; Civil Service)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRegional \u0026amp; Location-based Info\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTopic 1 (Location-based Facilities/Business Info)\u003c/p\u003e \u003cp\u003eTopic 20 (Public Facility Location Info)\u003c/p\u003e \u003cp\u003eTopic 21 (Institutional Research Performance Indicators)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e4.3. Semantic Topology and Spatial Distribution of Topics\u003c/h2\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e4\u003c/span\u003e illustrates the semantic topology and spatial distribution of representative topic clusters derived from public science and technology (S\u0026amp;T) data metadata. The two-dimensional embedding, generated using BERTopic and projected via UMAP, represents latent semantic relationships among datasets, where spatial proximity reflects higher semantic similarity. As emphasized in prior scientometric studies, such low-dimensional embeddings should not be interpreted as literal geometric distances but as relational representations that reveal the underlying conceptual organization of large textual corpora (B\u0026ouml;rner et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2003\u003c/span\u003e; Gokhberg et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe overall distribution reveals a densely populated central region primarily composed of clusters related to R\u0026amp;D project management, administrative coordination, and performance evaluation. This finding suggests that administrative and managerial metadata function as a semantic backbone within the public S\u0026amp;T data ecosystem. Similar patterns have been observed in science policy corpora, where governance-related concepts often occupy structurally central positions due to their integrative role across heterogeneous knowledge domains (Aria et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Waltman, \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). From a FAIR perspective, these centrally located clusters are strongly associated with Findability and Accessibility, as they frequently include persistent identifiers, project codes, institutional affiliations, and access-related attributes that facilitate data discovery and coordination (Wilkinson et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Jacobsen et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn contrast, clusters positioned toward the periphery of the embedding space exhibit more domain-specific semantic characteristics. Clusters associated with basic science and observational data\u0026mdash;such as meteorological measurements and bio-health datasets\u0026mdash;form relatively compact and distinct groupings. These clusters are characterized by the prevalence of raw variables, measurement units, and experimental context descriptors, indicating a stronger alignment with the \u003cb\u003eReusability\u003c/b\u003e dimension of FAIR. This pattern is consistent with empirical findings that raw scientific data tends to be semantically cohesive yet operationally isolated when structural metadata and common vocabularies are insufficiently standardized (Edwards et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Pasquetto et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eClusters related to research infrastructure, experimental environments, and standardization practices occupy intermediate positions between central administrative clusters and peripheral raw data clusters. This spatial configuration reflects their dual semantic role. On the one hand, these topics connect to managerial metadata through standardized classifications, process codes, and equipment identifiers. On the other hand, they support experimental reproducibility by documenting physical research settings and technical specifications. Prior studies on reproducibility and data stewardship emphasize that such infrastructural metadata constitutes a critical connective layer enabling cross-domain interoperability (Stodden et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Leonelli, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eMeanwhile, clusters associated with industrial support, technology transfer, and digital content appear in localized regions at moderate semantic distances from core research clusters. This pattern suggests that industry-facing datasets are conceptually adjacent to research outputs but remain partially decoupled from raw scientific data. This observation aligns with interpretations of Open Science as an extended ecosystem that includes policy and industry actors, while also highlighting persistent structural boundaries that limit seamless data reuse across institutional sectors (Ruijer \u0026amp; Meijer, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; OECD, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2021\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eTaken together, the semantic topology depicted in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e4\u003c/span\u003e demonstrates that FAIR-related attributes are unevenly distributed across topic clusters rather than uniformly embedded throughout the public S\u0026amp;T data landscape. The spatial configuration indicates that openness alone does not guarantee interoperability or reusability. Instead, these qualities emerge from the presence of standardized, machine-readable structural metadata that bridge semantically distant domains. This finding reinforces recent arguments in scientometrics and data studies that advancing Open Science requires not only expanding data disclosure but also strategically strengthening metadata infrastructures to enable large-scale integration, reuse, and cumulative knowledge production (Borgman, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2019b\u003c/span\u003e; Mons et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2017\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e4.4. Semantic Diagnosis of Open Science Connectivity based on FAIR Principles\u003c/h2\u003e \u003cp\u003eTo assess whether the semantic structures of the 30 derived topics contribute to the realization of an actual Open Science ecosystem, this study employed the FAIR principles proposed by Wilkinson et al. (\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2016\u003c/span\u003e) as an analytical framework. The FAIR principles\u0026mdash;Findability, Accessibility, Interoperability, and Reusability\u0026mdash;serve as essential prerequisites for data to function as scientific assets beyond the scope of mere information disclosure. The results of mapping the keywords and semantic structures extracted via topic modeling to these four criteria are as follows.\u003c/p\u003e \u003cp\u003eFirst, in terms of Findability, it was confirmed that public science and technology data encompass unique identifiers and rich metadata. In Topic 1 (Intellectual Property \u0026amp; Patent Status) and Topic 8 (Academic Literature \u0026amp; Bibliographic Info), unique identifiers such as 'Registration number,' 'Application number,' 'DOI,' 'Author,' and 'Signature' were derived as core keywords. This implies that data can be identified as unique objects, suggesting that an indexing foundation is established to allow researchers to precisely search for necessary resources amidst a vast sea of data. Furthermore, Topic 21 (Public Facility Location Info) includes spatial coordinate data like 'Latitude' and 'Longitude,' thereby extending the findability of GIS-based data into physical space.\u003c/p\u003e \u003cp\u003eSecond, regarding Accessibility, the data present clear access protocols and usage conditions. Keywords such as 'URL,' 'Link,' and 'Attachment file' appearing in Topic 11 (Online Board Activity Status) and Topic 19 (Digital Content Metadata) indicate that physical paths for users to access the actual data entities are explicitly specified within the metadata. Notably, keywords such as 'Paid,' 'Free,' and 'By target' included in Topic 13 (Tech Transfer \u0026amp; Corporate Support) reveal that information regarding data or technology usage rights and licensing is provided simultaneously. This proves that the opened data exists in a state where it can be actually acquired without legal or economic barriers, or under clearly stipulated conditions.\u003c/p\u003e \u003cp\u003eThird, concerning Interoperability, the adoption of machine-readable standard vocabularies and code systems was identified. The keywords 'Scientific name' and 'Common name' derived from Topic 3 (Biodiversity \u0026amp; Forest Resources) demonstrate adherence to international standard classification systems in biology, implying that biological data produced by different research institutions can be integrated based on species information. Additionally, the 'Process code' in Topic 14 (Manufacturing Process \u0026amp; Part Data) and the 'Classification code' in Topic 27 (Radio Management \u0026amp; Station Permit) corroborate that data are managed not as unstructured text but as standardized code values. Such compliance with standards is a critical element enabling data exchange and convergent analysis between heterogeneous systems (Janssen et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). However, the topics identified as 'Absence of column info' in Topic 23 and Topic 29 reveal that some datasets remain in a non-standardized state. This serves as a counter-example highlighting the urgency of advancing data quality management systems.\u003c/p\u003e \u003cp\u003eFourth, this study identified the most significant correlation with Open Science in the aspect of Reusability. Reusability is a core value determining whether data can be utilized as material for new research beyond its initial collection purpose (Murray-Rust, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2008\u003c/span\u003e). The analysis revealed that Topic 17 (Weather Observation \u0026amp; Aviation Info) and Topic 24 (Bio-health \u0026amp; Body Composition) contain observational variables themselves, such as 'Wind speed,' 'Wind direction,' 'Body fat mass,' and 'Water ratio,' rather than processed statistics. Furthermore, Topic 7 (Research Equipment \u0026amp; Specifications) and Topic 14 (Manufacturing Process \u0026amp; Part Data) provide detailed information on the physical environment and context in which research was conducted, including 'Model name,' 'Manufacturer,' 'Substrate size,' and 'Process name.' This supports the ideals of Open Science\u0026mdash;specifically research transparency and reproducibility\u0026mdash;by providing a foundation for third-party researchers to verify preceding studies or replicate experiments by reconstructing the same environment using the data (Munaf\u0026ograve; et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2017\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn conclusion, public science and technology data are evaluated to satisfy the four requirements of the FAIR principles at a semantic level. In particular, the structural identification of raw data and experimental context information, transcending administrative result reports, empirically demonstrates that the current data opening policy is evolving beyond simple transparency enhancement into an Open Science ecosystem for practical scientific value creation (OECD, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Vicente-Saez \u0026amp; Martinez-Fuentes, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e demonstrates that different semantic topic clusters align with distinct FAIR principles based on concrete metadata attributes rather than normative assumptions.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eEmpirical mapping of representative semantic topics to FAIR principles in public S\u0026amp;T data\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFAIR Principle\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCore Requirement\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMatched Topics\u003c/p\u003e \u003cp\u003e(Topic ID \u0026amp; Label)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eKey Empirical Evidence\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFindable\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePersistent identifiers and rich descriptive metadata\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTopic 1 (Intellectual Property);\u003c/p\u003e \u003cp\u003eTopic 8 (Bibliographic Information);\u003c/p\u003e \u003cp\u003eTopic 21 (Institutional Indicators)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026bull; Persistent identifiers: registration number, application number, DOI\u003c/p\u003e \u003cp\u003e\u0026bull; Structured bibliographic metadata enabling systematic discovery\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAccessible\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eClear access protocols and usage conditions\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTopic 11 (Advertising Market Statistics);\u003c/p\u003e \u003cp\u003eTopic 19 (Digital Content);\u003c/p\u003e \u003cp\u003eTopic 13 (Technology Transfer)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026bull; Access paths: URL, hyperlinks, attached files\u003c/p\u003e \u003cp\u003e\u0026bull; Explicit usage conditions: free, paid, target-specific licensing\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eInteroperable\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eStandardized vocabularies and machine-readable structures\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTopic 3 (ICT R\u0026amp;D and Bio-resources);\u003c/p\u003e \u003cp\u003eTopic 14 (Administrative Contact and Manufacturing Process);\u003c/p\u003e \u003cp\u003eTopic 27 (Broadcast Programming Data)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026bull; Controlled vocabularies: scientific names, common names based on bio-standards\u003c/p\u003e \u003cp\u003e\u0026bull; Standard classification and process codes enabling cross-dataset integration\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eReusable\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eProvision of raw data and contextual metadata\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTopic 17 (Research Agencies and Weather Data);\u003c/p\u003e \u003cp\u003eTopic 24 (R\u0026amp;D Workforce and Bio-health);\u003c/p\u003e \u003cp\u003eTopic 7 (Research Equipment);\u003c/p\u003e \u003cp\u003eTopic 13 (Manufacturing Process Data)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026bull; Raw variables: wind speed, wind direction, body fat mass, water ratio\u003c/p\u003e \u003cp\u003e\u0026bull; Contextual metadata: model name, manufacturer, substrate size, experimental conditions\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"5. DISCUSSION","content":"\u003cp\u003eThis study contributes to the growing body of Open Science and open data research by empirically examining the semantic and structural characteristics of public science and technology (S\u0026amp;T) data through large-scale metadata analysis. Unlike prior studies that primarily evaluate policy compliance or portal-level performance indicators, this research focuses on the latent semantic architecture embedded within unstructured metadata and its alignment with the FAIR principles. By employing BERTopic-based topic modeling and hierarchical clustering, the analysis reveals how public S\u0026amp;T data are conceptually organized across the R\u0026amp;D lifecycle and how structural metadata mediates their scientific usability.\u003c/p\u003e \u003cp\u003eThe findings demonstrate that public S\u0026amp;T data are semantically structured as an interconnected ecosystem rather than a collection of isolated datasets. The emergence of multiple topic clusters spanning research outcomes, infrastructure, industrial support, and raw observational data suggests that openness has increasingly extended upstream and downstream within the research process. This observation aligns with recent conceptualizations of Open Science as an infrastructural transformation rather than a narrow dissemination practice (Borgman, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2019a\u003c/span\u003e; Edwards et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). In particular, Edwards et al. (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2011\u003c/span\u003e) emphasize that data infrastructures shape what kinds of scientific practices become scalable, a perspective that helps explain why administrative and management-related metadata emerge as structurally central in the semantic topology observed in this study.\u003c/p\u003e \u003cp\u003eFrom a FAIR perspective, the semantic proximity between raw data topics and contextual metadata\u0026mdash;such as equipment specifications, experimental environments, and process descriptions\u0026mdash;indicates a growing orientation toward reproducibility and reuse. This supports recent arguments that reproducibility is not achieved through data availability alone, but through the availability of sufficiently rich contextual information that allows analytical reconstruction (Stodden et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Goodman et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). The presence of semantically dense clusters related to meteorological and bio-health data further highlights the role of public S\u0026amp;T data as reusable scientific inputs, echoing empirical findings that well-curated datasets generate cumulative value through secondary analysis and cross-domain recombination (Piwowar et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2018\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAt the same time, the results expose a critical structural imbalance. While semantically rich topics exist, several clusters exhibit ambiguous or weakly defined structural metadata, limiting their machine-actionable interoperability. This finding resonates with recent critiques that many open data initiatives remain \u0026ldquo;human-readable but machine-hostile,\u0026rdquo; thereby constraining large-scale computational reuse (Wilkinson et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Jacobsen et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Jacobsen et al. (\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) argue that without consistent schema definitions and standardized variable-level metadata, FAIR principles cannot be operationalized in practice\u0026mdash;a claim strongly supported by the unidentified and structurally fragile topic clusters identified in this analysis.\u003c/p\u003e \u003cp\u003eBeyond academic reuse, the semantic coupling between intellectual property, technology transfer, and industrial support topics underscores the innovation-oriented dimension of public S\u0026amp;T data. This configuration suggests that open research data function as boundary objects connecting public research and private-sector innovation systems. Recent studies on data-driven innovation emphasize that such spillovers depend less on openness per se than on interoperability and legal clarity (OECD, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). In this context, the study\u0026rsquo;s findings imply that deficiencies in structural metadata may inhibit not only scientific reuse but also downstream industrial valorization.\u003c/p\u003e \u003cp\u003eMethodologically, this study advances scientometric and open data research by demonstrating the analytical value of topic modeling grounded in semantic embeddings rather than frequency-based indicators alone. While previous studies have applied topic modeling to identify thematic orientations in public data portals, they often stop short of linking semantic patterns to infrastructural operability (Cho, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). By integrating semantic clustering with FAIR-oriented interpretation, this research bridges the gap between computational text analysis and data governance theory.\u003c/p\u003e \u003cp\u003eOverall, the findings suggest that future Open Science policies should shift emphasis from the quantitative expansion of datasets toward the qualitative strengthening of structural metadata. Mandating standardized, machine-readable schemas, persistent identifiers at the variable level, and explicit contextual documentation would significantly enhance interoperability and reusability. Such measures are increasingly recognized as prerequisites for sustainable Open Science ecosystems capable of supporting reproducibility, cross-sector collaboration, and cumulative knowledge production (Thibault et al., \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eDespite these contributions, this study is subject to several limitations. First, the empirical analysis was restricted to the South Korean Open Data Portal, which may limit the generalizability of the findings to other national contexts with different data governance models. Second, the semantic diagnosis relied primarily on textual and structural metadata without physically validating the internal quality or completeness of the actual data files (e.g., cell-level values), leaving the technical assessment of data integrity for future research. Finally, because the analysis focused on supply-side metadata provided by government agencies, it does not directly capture the demand-side user experience or the actual downstream reuse patterns of the datasets.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSummary of key empirical findings in relation to research questions\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eResearch Question\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAnalytical Focus\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eKey Results\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRQ1. What semantic roles does Open Data play within the discourse of Open Science in the S\u0026amp;T sector?\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSemantic clustering and topic interpretation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePublic S\u0026amp;T data form a multi-layered semantic structure spanning the entire R\u0026amp;D lifecycle, indicating that Open Data functions as a connective infrastructure rather than a collection of isolated datasets.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRQ2. To what extent does Open Data occupy a central or peripheral position in the semantic structure of Open Science?\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSpatial topology and semantic centrality\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTopics related to R\u0026amp;D project management and administrative metadata occupy central positions, acting as semantic backbones, while scientifically rich raw data topics remain peripheral, revealing structural asymmetries in data integration.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRQ3. How does the semantic positioning of Open Data relate to FAIR principles across domains?\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFAIR-oriented semantic mapping\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFAIR-aligned attributes are unevenly distributed across topics; interoperability and reusability depend on the presence of standardized structural metadata rather than openness alone.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e"},{"header":"6. CONCLUSION","content":"\u003cp\u003eThis study set out to examine how public science and technology (S\u0026amp;T) data are semantically structured and positioned within the evolving Open Science ecosystem, moving beyond conventional evaluations centered on policy compliance or mere data availability. By applying BERTopic-based topic modeling and hierarchical clustering to large-scale unstructured metadata from a public data portal, the analysis provided an empirical diagnosis of how Open Data in the S\u0026amp;T domain embodies core Open Science principles\u0026mdash;particularly transparency, reproducibility, and collaboration\u0026mdash;as articulated through the FAIR framework (Wilkinson et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe findings demonstrate that public S\u0026amp;T data form a multi-layered semantic structure encompassing the entire R\u0026amp;D lifecycle, rather than a fragmented or purely administrative collection of records. The identification of four core semantic axes\u0026mdash;Research Outcomes, Research Infrastructure, Industry Support, and Raw Data\u0026mdash;reveals that contemporary data openness extends beyond the disclosure of final outputs toward the inclusion of intermediate products and input resources generated throughout research processes. This structural configuration indicates a maturation of data-opening strategies toward ecosystem-level openness, rather than output-centric disclosure, aligning with the view that open data ecosystems must evolve to support diverse stakeholder interactions (Janssen et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Zuiderwijk et al., \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2014\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eImportantly, the study provides empirical evidence that public S\u0026amp;T data are semantically aligned with the principle of reproducibility. The presence of topics containing contextual information on experimental environments, equipment specifications, and process conditions suggests a transition from result-oriented data sharing to context-aware data provision. Such semantic characteristics enhance the potential for third-party verification, replication, and secondary analysis, thereby supporting cumulative knowledge production and the long-term sustainability of Open Science (Munaf\u0026ograve; et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Murray-Rust, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2008\u003c/span\u003e). This finding empirically validates the theoretical argument that data must be accompanied by provenance information to serve as epistemic resources (Leonelli, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAt the same time, this study highlights a critical limitation in current open data practices: the persistence of structural metadata deficiencies. While descriptive metadata is often sufficiently detailed, inconsistencies in structural metadata\u0026mdash;particularly non-standardized column names and schema designs embedded in raw datasets\u0026mdash;significantly undermine interoperability and computational reuse. This finding underscores that semantic richness alone is insufficient to realize the full benefits of Open Science if data operability remains constrained (Kubler et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). It reinforces the notion that without machine-actionable standards, data remains \"technically open\" but \"practically unusable\" for large-scale integration (Mons et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2017\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eBy shifting the analytical focus from metadata description to metadata operability, this study advances scientometric research on Open Data quality. The proposed semantic diagnostic framework illustrates how latent thematic structures and structural weaknesses can be jointly identified, offering a more nuanced understanding of data readiness for reuse than indicator-based assessments alone. In doing so, the study contributes methodologically by demonstrating the value of computational semantic analysis as a tool for evaluating the conceptual and infrastructural dimensions of Open Science (Grootendorst, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Vicente-Saez \u0026amp; Martinez-Fuentes, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2018\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eFrom a policy and governance perspective, the results suggest that future Open Science initiatives should prioritize qualitative refinement over further quantitative expansion. Specifically, policy interventions should mandate standardized, machine-readable structural metadata alongside descriptive documentation to reduce data linkage costs, enhance reproducibility, and facilitate cross-domain and cross-sectoral data integration. Strengthening metadata operability is therefore not merely a technical improvement, but a prerequisite for translating openness into sustained scientific innovation and socio-economic value (OECD, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2015\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eTaken together, this study reinforces the view that the effectiveness of Open Science depends not only on the volume or visibility of open data, but on its capacity to be semantically interpretable and structurally interoperable. By empirically diagnosing both dimensions, the study offers a foundation for future research and policy efforts aimed at building robust, reusable, and scalable Open Science infrastructures.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eJ.J. conceived the study, designed the methodology, collected and analyzed the data, and wrote the manuscript. The author read and approved the final manuscript.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003ePublicly available in a repository: The datasets analyzed during the current study are available in the Public Data Portal (Ministry of the Interior and Safety, Republic of Korea) at https://www.data.go.kr.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAria, M., Misuraca, M., \u0026amp; Spano, M. (2020). Mapping the evolution of social research and data science on 30 years of social indicators research. \u003cem\u003eSocial indicators research\u003c/em\u003e, \u003cem\u003e149\u003c/em\u003e(3), 803\u0026ndash;831.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBertot, J. C., Jaeger, P. T., \u0026amp; Grimes, J. M. (2010). Using ICTs to create a culture of transparency: E-government and social media as openness and anti-corruption tools for societies. \u003cem\u003eGovernment Information Quarterly\u003c/em\u003e, 27(3), 264\u0026ndash;271. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.giq.2010.03.001\u003c/span\u003e\u003cspan address=\"10.1016/j.giq.2010.03.001\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBlei, D. M., \u0026amp; Lafferty, J. D. (2006). Dynamic topic models. In \u003cem\u003eProceedings of the 23rd International Conference on Machine Learning\u003c/em\u003e (pp. 113\u0026ndash;120). ACM. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1145/1143844.1143859\u003c/span\u003e\u003cspan address=\"10.1145/1143844.1143859\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBlei, D. M., Ng, A. Y., \u0026amp; Jordan, M. I. (2003). Latent Dirichlet Allocation. \u003cem\u003eJournal of Machine Learning Research\u003c/em\u003e, 3(Jan), 993\u0026ndash;1022. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.5555/944919.944937\u003c/span\u003e\u003cspan address=\"10.5555/944919.944937\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBorgman, C. L. (2012). The conundrum of sharing research data. \u003cem\u003eJournal of the American Society for Information Science and Technology\u003c/em\u003e, 63(6), 1059\u0026ndash;1078. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/asi.22634\u003c/span\u003e\u003cspan address=\"10.1002/asi.22634\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBorgman, C. L. (2015). \u003cem\u003eBig Data, Little Data, No Data: Scholarship in the Networked World\u003c/em\u003e. MIT Press.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBorgman, C. L. (2019a). Open data, grey data, and stewardship: Universities at the privacy frontier. \u003cem\u003eBerkeley Technology Law Journal\u003c/em\u003e, 33(2), 365\u0026ndash;412.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBorgman, C. L. (2019b). The lives and after lives of data. \u003cem\u003eHarvard Data Science Review\u003c/em\u003e, 1(1). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1162/99608f92.9a36bdb6\u003c/span\u003e\u003cspan address=\"10.1162/99608f92.9a36bdb6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eB\u0026ouml;rner, K., Chen, C., \u0026amp; Boyack, K. W. (2003). Visualizing knowledge domains. \u003cem\u003eAnnual Review of Information Science and Technology\u003c/em\u003e, 37(1), 179\u0026ndash;255.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCampello, R. J. G. B., Moulavi, D., \u0026amp; Sander, J. (2013). Density-based clustering based on hierarchical density estimates. In \u003cem\u003ePacific-Asia Conference on Knowledge Discovery and Data Mining\u003c/em\u003e (pp. 160\u0026ndash;172). Springer. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-3-642-37456-2_14\u003c/span\u003e\u003cspan address=\"10.1007/978-3-642-37456-2_14\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCarbonell, J., \u0026amp; Goldstein, J. (1998). The use of MMR, diversity-based reranking for reordering documents and producing summaries. In \u003cem\u003eProceedings of the 21st ACM SIGIR Conference\u003c/em\u003e (pp. 335\u0026ndash;336). ACM. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1145/290941.291025\u003c/span\u003e\u003cspan address=\"10.1145/290941.291025\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCho, J. (2023). Comparative analysis of open government data topics and usability. \u003cem\u003eQuality \u0026amp; Quantity\u003c/em\u003e, \u003cem\u003e57\u003c/em\u003e(6), 5655\u0026ndash;5671.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDai, S. C. (2025). Conceptualizing Digital Democracy: A Policy-Oriented Approach Through Computational Text Analysis. \u003cem\u003ePolicy \u0026amp; Internet\u003c/em\u003e, \u003cem\u003e17\u003c/em\u003e(4), e70022.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEdwards, P. N., Mayernik, M. S., Batcheller, A. L., Bowker, G. C., \u0026amp; Borgman, C. L. (2011). Science friction: Data, metadata, and collaboration. \u003cem\u003eSocial Studies of Science\u003c/em\u003e, 41(5), 667\u0026ndash;690. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/0306312713491721\u003c/span\u003e\u003cspan address=\"10.1177/0306312713491721\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFanelli, D., et al. (2018). Is science really facing a reproducibility crisis? \u003cem\u003eProceedings of the National Academy of Sciences\u003c/em\u003e, 115(11), 2628\u0026ndash;2631.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFecher, B., \u0026amp; Friesike, S. (2014). Open Science: One Term, Five Schools of Thought. In S. Bartling \u0026amp; S. Friesike (Eds.), \u003cem\u003eOpening Science\u003c/em\u003e (pp. 17\u0026ndash;47). Springer. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-3-319-00026-8_2\u003c/span\u003e\u003cspan address=\"10.1007/978-3-319-00026-8_2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGoodman, S. N., Fanelli, D., \u0026amp; Ioannidis, J. P. A. (2016). What does research reproducibility mean? \u003cem\u003eScience Translational Medicine\u003c/em\u003e, 8(341), 341ps12. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1126/scitranslmed.aaf5027\u003c/span\u003e\u003cspan address=\"10.1126/scitranslmed.aaf5027\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGrootendorst, M. (2022). BERTopic: Neural topic modeling with a class-based TF-IDF procedure. \u003cem\u003earXiv preprint arXiv:2203.05794\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.48550/arXiv.2203.05794\u003c/span\u003e\u003cspan address=\"10.48550/arXiv.2203.05794\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGokhberg, L., Meissner, D., \u0026amp; Kuzminov, I. (2023). What semantic analysis can tell us about long term trends in the global STI policy agenda. \u003cem\u003eThe Journal of Technology Transfer\u003c/em\u003e, \u003cem\u003e48\u003c/em\u003e(6), 2249\u0026ndash;2277.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHecker, S., Haklay, M., Bowser, A., Makuch, Z., Vogel, J., \u0026amp; Bonn, A. (2018). \u003cem\u003eCitizen Science: Innovation in Open Science, Society and Policy\u003c/em\u003e. UCL Press. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.14324/111.9781787352339\u003c/span\u003e\u003cspan address=\"10.14324/111.9781787352339\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJacobsen, A., de Miranda Azevedo, R., Juty, N., Batista, D., Coles, S., Cornet, R., \u0026hellip; Mons, B. (2020). FAIR principles: Interpretations and implementation considerations. \u003cem\u003eData Intelligence\u003c/em\u003e, 2(1\u0026ndash;2), 10\u0026ndash;29. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1162/dint_r_00024\u003c/span\u003e\u003cspan address=\"10.1162/dint_r_00024\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJanssen, M., Charalabidis, Y., \u0026amp; Zuiderwijk, A. (2012). Benefits, adoption barriers and myths of open data and open government. \u003cem\u003eInformation Systems Management\u003c/em\u003e, 29(4), 258\u0026ndash;268. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/10580530.2012.716740\u003c/span\u003e\u003cspan address=\"10.1080/10580530.2012.716740\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKitchin, R. (2014). \u003cem\u003eThe Data Revolution: Big Data, Open Data, Data Infrastructures and Their Consequences\u003c/em\u003e. Sage.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKubler, S., Robert, J., Neumaier, S., Umbrich, J., \u0026amp; Le Traon, Y. (2018). Open data portal quality comparison using AHP. \u003cem\u003eGovernment Information Quarterly\u003c/em\u003e, 35(2), 13\u0026ndash;29. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.giq.2018.02.001\u003c/span\u003e\u003cspan address=\"10.1016/j.giq.2018.02.001\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKučera, J., Chlapek, D., \u0026amp; Nečask\u0026yacute;, M. (2013). Quality assessment of open government data. In \u003cem\u003eInternational Conference on Electronic Government\u003c/em\u003e (pp. 351\u0026ndash;359). Springer. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-3-642-40160-2_28\u003c/span\u003e\u003cspan address=\"10.1007/978-3-642-40160-2_28\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee, J. (2020). Demand forecasting of public data using keyword network analysis. \u003cem\u003eInformatization Policy\u003c/em\u003e, 27(4), 24\u0026ndash;46. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.22693/NIAIP.2020.27.4.024\u003c/span\u003e\u003cspan address=\"10.22693/NIAIP.2020.27.4.024\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLeonelli, S. (2016). \u003cem\u003eData-Centric Biology: A Philosophical Study\u003c/em\u003e. University of Chicago Press.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcInnes, L., Healy, J., \u0026amp; Melville, J. (2018). UMAP: Uniform Manifold Approximation and Projection for dimension reduction. \u003cem\u003earXiv preprint arXiv:1802.03426\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.48550/arXiv.1802.03426\u003c/span\u003e\u003cspan address=\"10.48550/arXiv.1802.03426\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMons, B., Neylon, C., Velterop, J., Dumontier, M., da Silva Santos, L. O. B., \u0026amp; Wilkinson, M. D. (2017). Cloudy, increasingly FAIR; revisiting the FAIR data guiding principles for the European Open Science Cloud. \u003cem\u003eInformation Services \u0026amp; Use\u003c/em\u003e, 37(1), 49\u0026ndash;56. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3233/ISU-170824\u003c/span\u003e\u003cspan address=\"10.3233/ISU-170824\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMunaf\u0026ograve;, M. R., Nosek, B. A., Bishop, D. V., Button, K. S., Chambers, C. D., Du Sert, N. P., \u0026hellip; Ioannidis, J. P. (2017). A Manifesto for Reproducible Science. \u003cem\u003eNature Human Behaviour\u003c/em\u003e, 1(1), 0021. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41562-016-0021\u003c/span\u003e\u003cspan address=\"10.1038/s41562-016-0021\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMurray-Rust, P. (2008). Open Data in Science. \u003cem\u003eSerials Review\u003c/em\u003e, 34(1), 52\u0026ndash;64. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.serrev.2008.01.001\u003c/span\u003e\u003cspan address=\"10.1016/j.serrev.2008.01.001\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOECD. (2015). \u003cem\u003eData-Driven Innovation: Big Data for Growth and Well-Being\u003c/em\u003e. OECD Publishing. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1787/9789264229358-en\u003c/span\u003e\u003cspan address=\"10.1787/9789264229358-en\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOECD. (2021). \u003cem\u003eEnhancing Access to and Sharing of Data: Reconciling Risks and Benefits for Data Re-use across Societies\u003c/em\u003e. OECD Publishing. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1787/276aaca8-en\u003c/span\u003e\u003cspan address=\"10.1787/276aaca8-en\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePasquetto, I. V., Borgman, C. L., \u0026amp; Wofford, M. F. (2019). Uses and reuses of scientific data: The data creators' advantage. \u003cem\u003eProceedings of the National Academy of Sciences\u003c/em\u003e, 116(10), 4108\u0026ndash;4117. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1073/pnas.1804247116\u003c/span\u003e\u003cspan address=\"10.1073/pnas.1804247116\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePiwowar, H. A., \u0026amp; Vision, T. J. (2013). Data reuse and the open data citation advantage. \u003cem\u003ePeerJ\u003c/em\u003e, 1, e175. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.7717/peerj.175\u003c/span\u003e\u003cspan address=\"10.7717/peerj.175\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePiwowar, H. A., Priem, J., Larivi\u0026egrave;re, V., Alperin, J. P., Matthias, L., Norlander, B., \u0026hellip; Haustein, S. (2018). The state of OA: A large-scale analysis of the prevalence and impact of Open Access articles. \u003cem\u003ePeerJ\u003c/em\u003e, 6, e4375. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.7717/peerj.4375\u003c/span\u003e\u003cspan address=\"10.7717/peerj.4375\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQiang, J., Qian, Z., Li, Y., Yuan, Y., \u0026amp; Wu, X. (2020). Short text topic modeling techniques, applications, and performance: A survey. \u003cem\u003eIEEE Transactions on Knowledge and Data Engineering\u003c/em\u003e, 34(3), 1427\u0026ndash;1445. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/TKDE.2020.2999485\u003c/span\u003e\u003cspan address=\"10.1109/TKDE.2020.2999485\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eReimers, N., \u0026amp; Gurevych, I. (2019). Sentence-BERT: Sentence embeddings using Siamese BERT-networks. In \u003cem\u003eProceedings of EMNLP-IJCNLP 2019\u003c/em\u003e (pp. 3982\u0026ndash;3992). ACL. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.18653/v1/D19-1410\u003c/span\u003e\u003cspan address=\"10.18653/v1/D19-1410\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRuijer, E., \u0026amp; Meijer, A. (2020). Open government data as an innovation process: Lessons from a living lab experiment. \u003cem\u003ePublic Performance \u0026amp; Management Review\u003c/em\u003e, \u003cem\u003e43\u003c/em\u003e(3), 613\u0026ndash;635.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStodden, V., McNutt, M., Bailey, D. H., Deelman, E., Gil, Y., Hanson, B., \u0026hellip; Taufer, M. (2016). Enhancing reproducibility for computational methods. \u003cem\u003eScience\u003c/em\u003e, 354(6317), 1240\u0026ndash;1241. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1126/science.aah6168\u003c/span\u003e\u003cspan address=\"10.1126/science.aah6168\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTenopir, C., Dalton, E. D., Allard, S., Frame, M., Pjesivac, I., Birch, B., \u0026hellip; Klentzens, K. (2011). Data sharing by scientists: practices and perceptions. \u003cem\u003ePLoS One\u003c/em\u003e, 6(6), e21101. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1371/journal.pone.0021101\u003c/span\u003e\u003cspan address=\"10.1371/journal.pone.0021101\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThibault, R. T., Amaral, O. B., Argolo, F., Bandrowski, A. E., Davidson, A. R., \u0026amp; Drude, N. I. (2023). Open Science 2.0: Towards a truly collaborative research ecosystem. \u003cem\u003ePLoS Biology\u003c/em\u003e, \u003cem\u003e21\u003c/em\u003e(10), e3002362. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1371/journal.pbio.3002362\u003c/span\u003e\u003cspan address=\"10.1371/journal.pbio.3002362\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUbaldi, B. (2013). Open government data: Towards empirical analysis of open government data initiatives. \u003cem\u003eOECD Working Papers on Public Governance\u003c/em\u003e, (22). OECD Publishing. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1787/5k46bj4f03s7-en\u003c/span\u003e\u003cspan address=\"10.1787/5k46bj4f03s7-en\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVicente-Saez, R., \u0026amp; Martinez-Fuentes, C. (2018). Open Science now: A systematic literature review. \u003cem\u003eJournal of Business Research\u003c/em\u003e, 88, 428\u0026ndash;436. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jbusres.2017.12.043\u003c/span\u003e\u003cspan address=\"10.1016/j.jbusres.2017.12.043\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWalker, R. M., Chandra, Y., Zhang, J., \u0026amp; van Witteloostuijn, A. (2019). Topic modeling the research-practice gap in public administration. \u003cem\u003ePublic administration review\u003c/em\u003e, \u003cem\u003e79\u003c/em\u003e(6), 931\u0026ndash;937.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWaltman, L. (2016). A review of the literature on citation impact indicators. \u003cem\u003eJournal of Informetrics\u003c/em\u003e, 10(2), 365\u0026ndash;391.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWilkinson, M. D., Dumontier, M., Aalbersberg, I. J., Appleton, G., Axton, M., Baak, A., \u0026hellip; Mons, B. (2016). The FAIR Guiding Principles for scientific data management and stewardship. \u003cem\u003eScientific Data\u003c/em\u003e, 3, 160018. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/sdata.2016.18\u003c/span\u003e\u003cspan address=\"10.1038/sdata.2016.18\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZuiderwijk, A., \u0026amp; Janssen, M. (2014). Open data policies, their implementation and impact. \u003cem\u003eGovernment Information Quarterly\u003c/em\u003e, 31(1), 17\u0026ndash;29. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.giq.2013.04.003\u003c/span\u003e\u003cspan address=\"10.1016/j.giq.2013.04.003\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZuiderwijk, A., Janssen, M., \u0026amp; Davis, C. (2014). Innovation with open data: Essential elements of open data ecosystems. \u003cem\u003eInformation Polity\u003c/em\u003e, 19(1, 2), 17\u0026ndash;33. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3233/IP-140329\u003c/span\u003e\u003cspan address=\"10.3233/IP-140329\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Open Science, Open Data, Semantic Analysis, Scientometrics, FAIR Principles, Science and Technology Policy","lastPublishedDoi":"10.21203/rs.3.rs-8440351/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8440351/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eOpen Science has emerged as a central paradigm in contemporary science and technology (S\u0026amp;T) policy, with Open Data widely regarded as one of its core components. Despite this prominence, limited empirical attention has been paid to whether Open Data occupies a structurally meaningful position within the semantic architecture of Open Science discourse. This study conducts a computational semantic analysis of public S\u0026amp;T data-related documents to diagnose the conceptual relationship between Open Data and Open Science. Using BERTopic-based modeling and hierarchical clustering, we examine how Open Data is positioned within the broader Open Science discourse, focusing on its centrality, proximity to key Open Science concepts, and alignment with FAIR principles. The results reveal that while Open Data is frequently referenced, it exhibits a distinct core-periphery structure: administrative and management-oriented metadata occupy a central semantic position, whereas scientifically rich raw data tend to remain on the periphery. The structural analysis further indicates that the semantic integration of Open Data remains uneven across domains, suggesting a partial decoupling between policy expectations and conceptual implementation. By providing a semantic diagnosis of Open Data within Open Science discourse, this study contributes to scientometric research by offering a structural perspective on how foundational concepts of Open Science are articulated and operationalized in practice. The findings highlight the need to move beyond declarative commitments toward a more conceptually integrated understanding of Open Data in the evolution of Open Science.\u003c/p\u003e","manuscriptTitle":"From Open Data to Open Science?: A Semantic Diagnosis of Public Science and Technology Data","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-29 16:57:13","doi":"10.21203/rs.3.rs-8440351/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"82f9e75a-3ae9-42a7-baf7-8393f60b483b","owner":[],"postedDate":"January 29th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-01-29T16:57:13+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-29 16:57:13","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8440351","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8440351","identity":"rs-8440351","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}