Patent Landscape of Generative Networks: A Data-Driven Examination of Essential Concepts Convergence Opportunities and Technological Trends Through Patent Analysis

Patent Landscape of Generative Networks: A Data-Driven Examination of Essential Concepts Convergence Opportunities and Technological Trends Through Patent Analysis | 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 Patent Landscape of Generative Networks: A Data-Driven Examination of Essential Concepts Convergence Opportunities and Technological Trends Through Patent Analysis Konstantinos Charmanas This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9446615/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 Technological novelties mark the rise of new ideas, impacting both scientific tools and industrial products, and can be monitored through public resources. Patents constitute a resource for tracking essential and breakthrough innovations to evaluate their overall development in the market. Generative Artificial Intelligence (AI) is one of these breakthroughs, offering capabilities that facilitate knowledge acquisition, writing, and image processing among many common and professional tasks. This study examines technologies related to generative networks using international patent documents from the Patent Lens Database. The Cooperative Patent Classification (CPC) system was the basis for identifying associated types of technologies and their interconnections through cluster analysis and convergence networks, while other metadata were also evaluated to address major trends and geographical patterns. CPC classifications show that generative networks are primarily combined with image processing techniques, data analysis pipelines, and computational models, while also finding implementations in other indirectly related fields concerning electric vehicles, healthcare, and business tasks. Technology convergence indicators showed multiple potential pathways that highlight the interconnections between different technologies and potential use cases. Patent application dates suggest the emergence of methods and applications concerning power efficiency, climate mitigation, measuring and testing, IoT, and graphical data reading, while the major concepts of data processing, image pattern recognition, and business implementations were classified as arising too. Evaluations regarding patent jurisdictions indicate the existence of experts and prevalent investors operating in distinct developed countries, whereas country-wise preferences are confirmed. Altogether, this study provides insights into major technological directions related to generative networks, demonstrates convergence opportunities, highlights recent trends, and addresses geographical patterns. Generative AI Patent Analysis Statistical Analysis Clustering Convergence Networks Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 1. Introduction Technological patents reflect the exclusive rights granted for valuable inventions, whose characteristics and patterns are protected for competitive and strategic reasons. The potential use, reproduction, or any kind of exploitation of an invention by third parties is successfully restricted so that the owners are the only ones benefiting from their offerings. Patents signify the importance of distinct inventions, which can cover breakthroughs improving the value of products, systems, and services or even the evaluation of a whole organization. One major technological field characterized by a massive rise and individual breakthroughs is generative Artificial Intelligence (AI), which primarily covers technologies for image, video, and natural language processing or generation [1]. These offerings highlighted the emergence of technologies in various areas of interest, including education [2], healthcare [3], and manufacturing [4]. While there is considerable research through scientific [1] and grey literature [5] reviews, patent data constitute a less popular source for overviewing industrial and research activities. Nonetheless, researchers have already shifted their attention towards patents related to AI, covering AI in general [6-10] and specific subjects, such as digital pathology [11], electric vehicles [12], and healthcare [13]. The broader term used to label studies dedicated to examining and summarizing patent data is called patent analysis, where some common objectives include trend and competitor analysis, technology forecasting, and assessing patent quality [14]. In the case of generative AI, research from Bach et al. [15] constitutes a preliminary patent analysis focusing on the Cooperative Patent Classification (CPC 1 ) scheme and the technical content of patent documents. Their goal was to search for technologies co-occurring with generative AI, hence providing valuable insights into prevalent subjects. By considering the rise of generative AI and valuable information included in patent data, the current study builds upon the outcomes produced by previous research by investigating technology-based patterns related to generative networks through global patent analysis. The primary Research Objectives (ROs) were the following: RO1 : Address major technological directions and convergence opportunities RO2 : Find emerging and declining technological concepts RO3 : Evaluate geographical patterns and differences To cover these objectives, occurrence and co-occurrence statistics, cluster analysis, and association rules constituted the main methods employed, while the Patent Lens Database 2 and the CPC scheme were the primary sources of information. In fact, patent CPC classifications were investigated as the ground truth to explore technology frequencies and patterns. To be more precise, RO1 was addressed through cluster analysis and association rules, where strong associations led to the development of a convergence network. In combination with cluster properties and CPC classifications, major trends were evaluated through patent application dates, while geographical patterns were assessed based on patent jurisdictions. In total, this study fills an existing gap regarding the patent activity surrounding generative networks via the provision of information regarding major and trending technologies and the insights concerning geographical and technological convergence. This information can contribute to future projects by offering guidelines about essential technologies considered by large companies and about more specific opportunities, adapting to the needs of a potential researcher, as reflected by geographical opportunities and convergence options of distinct applications of interest with other technological domains. The following section discusses the contribution and themes covered by previous patent analysis studies, focusing especially on papers dedicated to patents related to AI (Section 2). Section 3 presents the methodology employed to cover the ROs, while the following section includes the main outcomes of the analysis (Section 4). The paper concludes with Section 5 by discussing concluding remarks and recommendations for future research. 2. Related Work Patent analysis is the review of patented inventions covering any technological field and related contributors that shape the existing knowledge background and status of the field. Abbas et al. [14] describe a generic workflow for patent analysis and suggest that patent documents contain valuable information for research purposes, including technical characteristics (title, claims, abstract, description, patent classifications) and associated metadata such as patent inventors, applicants, and citations. After establishing structured patent data, researchers can evaluate data patterns to cover various objectives and challenges that could offer valuable insights about the current and future status of a technological field. In our previous research [16, 17], we also investigated properties of this type to identify primary and trending-declining topics and major contributors (inventors and applicants), as well as assess patent value through citation networks. The ideas and methodologies were inspired by previous research that set the basis for extracting patterns through appropriate indicators [18, 19], while many other research studies have previously proposed related approaches for capturing semantics [20], analyzing citation networks [21], competitor analysis [22], patent classifications [23], and patent value assessment [24]. A later bibliometric analysis of the literature [25] indicates that patent analysis is widely employed to investigate diverging technological areas of interest, as suggested by the journals publishing relevant material. These journals may range from innovation science, R&D, and business topics to computer science, e.g., information processing and expert systems. Based on their analysis, some more domain-specific themes related to engineering and energy were among the most frequent fields as well. Furthermore, AI methods have also been implemented in patent analysis frameworks for multiple relevant tasks. Lin and Chou [26] showed that AI finds implementations in tasks beyond predictive mechanisms (trend prediction and value assessment), including the evaluation of topics, emerging technologies, network graphs, and firm activities. Through a review dedicated to pretrained and large language models, Shomee et al. [27] highlight the value of these state-of-the-art models in patent analysis for the tasks of patent classification, quality-value analysis, and text generation. This information shows that AI can be used to address a plethora of tasks covering standard patent analysis scopes and objectives. Since the current study focuses on patents related to generative networks, it is worth discussing key insights and approaches from previous studies exploring patents associated with the general domain of AI. Fuji and Managi [6] were among the first to discuss patent activity surrounding AI, already identifying the involvement of major companies and the rapid activity surrounding relevant technologies. In the investigated timeline, the analysis showed that the attention was shifting from biological and knowledge-based models to mathematical and other AI technologies. Through text mining, Abadi and Pecht [7] later showed that fields related to neural networks, machine learning, and pattern recognition, along with specific algorithms like decision trees and support vector machines, were widely employed to develop technologies with AI capabilities. Through the CPC scheme, they also pointed out that techniques concerning digital data processing and transmission, as well as image and speech analysis, constitute some of the major associated methodologies. In addition, some indirectly related fields like pictorial and telephonic communication, business, and healthcare technologies were also apparent. Later research case studies also support the findings discussed above, confirming the rise of AI technologies whilst investigating other aspects as well, like country-level evaluations [8, 9], technology network analysis [9], and patent classification [10]. Among the latest technological fields, these studies also address the inclusion of natural language processing, hardware, computer vision, and telecommunication technologies in patents associated with AI. To find the use cases of AI in specific sciences, other studies have also dedicated their research to healthcare-related patents and subjects, showing the coverage of AI technologies in ophthalmology, oncology, radiology, cardiology, and general-purpose surgeries [13]. Similarly, Lee [12] discussed a plethora of different AI methods contributing to the various utilities of electric vehicles, with deep learning architectures, clustering algorithms, and decision trees being among the most frequent ones. In general, there are multiple studies highlighting the value of AI in diverging domains, including climate change [28], automation technologies [29], and sports [30]. Regarding generative AI, Bach et al. [15] showed that the patent activity surrounding technologies of this type had a dramatic rise around 2023 and 2024, indicating the tendency of major enterprises to invest in this breakthrough technological field. Through content-based analysis, they also revealed some relevant themes mentioned within patents of this type, like natural language processing, response generation, and training machine learning models. By further exploring CPC patent classifications, they also addressed some prevalent types of technologies concerning generative networks, machine learning, autoencoders, and adversarial learning. In total, our study builds upon these findings and provides insights into technology convergence and trend analysis while, at the same time, exploring geographical patterns. By investigating CPC classifications, the analysis concentrates on well-established technological fields and not on textual descriptions that may lead to irrelevant and repetitive content or non-interpretable topics. On top of that, both frequent and rare classifications are evaluated, while a cluster analysis method is also employed to provide collective information and cover the entirety of the observed classifications. 3. Methodology The current section demonstrates the methodology that led to the extraction of valuable insights from patent data related to generative networks. The first part of the methodology is the process of patent collection and preprocessing (Section 3.1), while the rest of the sections were organized based on the introduced ROs. Accordingly, Section 3.2 presents the methods followed to establish technology clusters and assess technology convergence opportunities (RO1). A trend analysis pipeline is covered in Section 3.3 (RO2), while Section 3.4 describes the statistical approach used to associate geographical information with technologies and detect patterns (RO3). Figure 1 provides an overview of the study. 3.1. Patent Collection and Preprocessing Initially, the Patent Lens Database was selected as the primary source of information, as it gathers global patent data and provides access to multiple resourceful metadata, such as application dates, patent classifications, content information (title, abstract), and jurisdictions. After searching the CPC scheme, the only encoding that is directly associated with generative networks is the CPC subgroup G06N3/0475, described as Generative networks . This subgroup falls under the CPC subgroups G06N3/03, entitled Neural networks , and G06N3/04, entitled Architecture, e.g., interconnection topology . In this case, the CPC subgroup G06N3/03 is the superconcept of G06N3/04. In general, the CPC scheme is organized into multiple hierarchy levels where the encoding of generative networks (G06N3/0475) can be disassembled as follows: G (Section), 06 (Class), N (Subclass), 3 (Group), 0475 (Subgroup). This search strategy led to the collection of 44230 patent applications and granted patents. However, patent documents (documents referring to applications, granted patents, etc.) may have overlaps due to patent renewal procedures, innovation variations and versions, as well as partial or complete technology expansions. For this reason, an approach towards merging observations covering similar content was considered to avoid data replication. Databases track the relationships between patent documents through simple and extended patent families. A simple patent family includes patents covering the same invention, while an extended family includes patents with linked inventions covering similar content. Based on the patent collections formed by extended families, a final dataset of 21048 extended patent families was established by merging related documents based on specific rules. The family sizes of the patents indicated that the inventions covered by the collected patents were relatively new, with a median family size equal to 1 and a third quartile (Q3) equal to 2, while there were some cases where the family size was extremely large (maximum value equal to 668). These observations show that the proposed approach of merging similar inventions was vital towards the validity of this study, as there were large patent families that could affect the outcomes significantly. In the current study, each family was associated with a CPC subgroup and a patent jurisdiction when at least one of its documents was associated with them. The information about the technologies covering each patent family was gathered based on CPC classifications instead of textual information because the various patent offices use different languages in textual information (titles and abstracts), while the CPC scheme is used globally. When it comes to date information, the earliest application dates of the retrieved documents were stored per family. To offer a solution that considers both the relationships between CPC encodings (avoid overlaps) and the patterns from different encodings, the CPC subclasses were investigated instead of the CPC subgroups. As a result, each patent family was associated with a CPC subclass with count indicators, according to the total number of CPC subgroups that fall under the subclass and were assigned to the patent family, and binary indicators denoting relevance. It should be noted that the CPC subgroup G06N3/0475 was excluded from the analysis, as it was used to retrieve relevant patents, while 309 CPC subclasses were observed in total. 3.2. Analyzing Technology Co-occurrence Patterns Overall, the primary focus of the current study is the investigation of patterns surrounding generative networks, as expressed by valuable inventions covered by patents. In this context, technology patterns were investigated through clustering algorithms (Section 3.2.1) and co-occurrence networks (Section 3.2.2). Multivariate analysis was used to cluster together multiple types of technologies based on CPC frequencies, hence finding the general technological directions associated with generative networks. Co-occurrence networks can be seen as a different but complementary approach, evaluating the convergence between pairs of technologies instead of groups, hence establishing convergence networks. Technology convergence is a valuable source of information towards identifying and monitoring emerging opportunities in the market [ 31 ], which in this case are measured based on the total interactions between two different types of technologies in patent data. 3.2.1. Technology Clustering Clustering algorithms usually rely on feature (data points) or similarity matrices. Accordingly, a Patent to CPC Matrix (PCM) was created, where each cell denoted the frequency of a CPC subclass (column) in a patent family (row). Similar to our previous research [ 17 , 32 ], a variation of Non-negative Matrix Factorization [ 33 , 34 ], described as Sparse Non-negative Matrix Factorization (SNMF) in this case [ 35 ], was employed to find technology clusters via the NIMFA Python library [ 36 ]. NMF models decompose an initial matrix of \(\:m\) features (rows) and \(\:n\) observations (columns) described as \(\:V\) ( \(\:m\:\times\:n\) ), inverted PCM in this case, into two lower-rank matrices reflecting the associations of \(\:r\) (rank) latent features (or factors) with the initial features and observations, usually denoted as \(\:W\) ( \(\:m\:\times\:r\) ) and \(\:H\) ( \(\:r\:\times\:n\) ), respectively. The general objective of NMF models is to approximate the matrix \(\:V\) as the product of \(\:W\) and \(\:H\) (1). $$\:{V}^{m\times\:n}=\:{W}^{m\times\:r}{H}^{r\times\:n}$$ 1 The benefit of SNMF compared to other variations is that it enforces sparse factors, meaning that each latent factor is expected to rely only on a few initial features, leading to more interpretable and distinct factors (technology clusters). In contrast to univariate analysis, this approach facilitates the collective investigation and summarization of all 309 CPC subclasses through data structures that are less complex than PCM, hence enabling pattern recognition and the inclusion of all information about patent classifications. A vital part of the methodology was the selection of an appropriate rank (number of factors), where the approximated low-dimensional matrices ( \(\:W\) and \(\:H\) ) both capture the variability of the initial matrix ( \(\:V\) ) and help summarize data patterns and feature correlations. In this case, the variability explained by each SNMF model and the topic divergence [ 37 ] between the extracted factors were selected to find the optimal rank, hence leading to factors with divergent compositions that also capture a high proportion of the available information. The first metric is calculated as the percentage of unique top features between all factors, where the top 3 and top 5 features per factor were considered. For example, given a model with two factors and the top 5 features, a topic divergence evaluation equal to 0.9 would mean that only a single feature appeared in the top features of both factors. After finding an appropriate rank, the weights stored in matrix \(\:W\) were evaluated to find the technical content covered by each technology cluster, while matrix \(\:H\) was considered as the baseline information towards associating technology clusters with jurisdictions and date variables. 3.2.2. Constructing Convergence Network A convergence network is a visual representation of strong associations between two technologies, showing their potential in individual innovations. Strong associations between two observations were evaluated using association rules considering the number of patents including both CPC subclasses at least once as a function of the number of patents including either one. Thus, the current study investigates both CPC frequencies (Section 3.2.1) and binary co-occurrences (Section 3.2.2) to address any potential limitations of each approach. Association rule mining is a wide family of data mining methods used to measure the relationships between two items, where one of the essential concepts is described as confidence. Given items \(\:X\) and \(\:Y\) , the confidence of \(\:X->Y\) measures the probability of \(\:Y\) , usually described as consequent, occurring in the observations containing \(\:X\) , usually described as antecedent [ 38 ]. This method measures the one-way dependency between the two items, where one is considered the antecedent and the other the consequent. To establish a single indicator, an alternative method described as the Inclusion Index (II) was employed [ 39 ]. II is the maximum probability between \(\:X->Y\) and \(\:Y->X\) , $$\:{II}_{xy}=\:\frac{P\left(XY\right)}{min\left(P\left(X\right),P\left(Y\right)\right)}\:$$ 2 where \(\:\text{P}\left(\text{X}\text{Y}\right)\) is the probability of \(\:X\) and \(\:Y\) co-occurring in a dataset while \(\:\text{P}\left(\text{X}\right)\) and \(\:\text{P}\left(\text{Y}\right)\) are the probabilities of \(\:X\) and \(\:Y\) occurring in the dataset, respectively. In the context of the current study, the benefit of II against other options is that this method also addresses the significance of one major technology to other minor fields, thus finding both one-way and two-way convergence opportunities. One-way means that the confidence of either \(\:X->Y\) and \(\:Y->X\) is high, while two-way convergence means that the confidence of both \(\:X->Y\) and \(\:Y->X\) are high. As a result, high II evaluations can lead to the identification of technology convergence without the need to define antecedents and consequents. For example, generative networks in combination with image analysis techniques may be significant in a variety of technological fields like medical diagnosis, vehicle control systems, pictorial communication, and material examination. In this case, II will respond to high evaluations in individual cases where image analysis techniques are frequent, while two-way approaches would not assess their co-frequencies as significant since these techniques apply to many independent domains. To establish convergence networks, a threshold of 0.4 II evaluations was set to denote strong associations indicating technology convergence between CPC subclasses. Also, only CPC subclasses that occurred in at least 20 patent families (close to 0.1%) were evaluated to avoid random events. 3.3. Finding Technological Trends Date information can be seen as a ground truth basis towards investigating the evolution and initial disclosure of ideas and implementations associated with a patent dataset to ultimately categorize technology directions as arising, essential, or declining. The earliest application dates of the established patent families were considered to identify arising and declining areas of interest. Starting from clusters, the frequency of each cluster was calculated per year along with two indicators used to evaluate their annual presence, based on the matrix \(\:H\) . First, the yearly cluster frequency was calculated. Since the analysis is based on the earliest application dates related to patent families, yearly cluster frequency is related to the disclosure of new inventions per year and the inclusion of technologies covered by each cluster, hence making this information appropriate for evaluating the overall evolution of a technology cluster annually. The first trend indicator measures the relative frequency of each cluster per patent and year, denoted as yearly patent share. The second indicator measures the relative frequency of a field per year with respect to the total frequency of that field in the investigated patent dataset, denoted as yearly technology share. These two indicators were used to identify the breakthrough years and the overall development of the various technologies over time to categorize them as arising or declining. Going beyond clusters, individual CPC subclasses that occurred in at least 20 patent families were also examined to provide a more thorough view of emerging technologies. To evaluate trending technologies, the frequencies of different CPC subclasses before and after 2023 were investigated and compared. The selected break point was marked by the launch of ChatGPT and other generative AI tools in late 2022, which undoubtedly changed the landscape of organizational procedures and everyday practices by offering new capabilities to end-users. An approach proposed by Charmanas et al. [ 40 ] was used to measure self-growth, where the number of topic assignments was replaced with total frequencies of CPC subclasses within a timeframe. In this case, the approach measures each technology’s growth after 2023 with respect to its overall frequency prior to this year. A self-growth equal to 2 would mean that the total frequency of the examined CPC subclass increased by 200% after 2023. 3.4. Exploring Geographical Patterns Individuals and organizations usually operate in distinct countries and regions rather than globally, meaning that they seek protection for their inventions through patents in the appropriate patent offices. Patent data offer information about the jurisdictions of each patent, which can be used to evaluate regional patterns. This information can be used to address the technological status and differences between various countries. Indicators of this type correspond to enterprise and academic activities that can be reviewed for standard patent analysis scopes like competitor analysis and strategic planning. This study investigates geographical patterns through technology clusters and major jurisdictions recorded for each patent. Each patent family was associated with a jurisdiction when at least one related patent application was filed to cover the invention in the corresponding country or group of countries, i.e., the European Patent Office (EPO 3 ). Then, the co-frequencies between each jurisdiction and cluster were calculated to compare the various countries according to technology patterns as well as find the most relevant to each cluster. The co-frequencies were standardized based on the total frequency of each cluster, providing ratios instead of absolute frequencies. An evaluation equal to 0.3 would mean that 30% of the investigated cluster is associated with the investigated jurisdiction. 4. Results and Discussion Section 4 includes the main findings of this study, addressing the ROs introduced in Section 1 with the use of the methods presented previously (Section 3). Sections 4.1 and 4.2 cover the first RO by investigating technology clusters and technology convergence opportunities, respectively. Furthermore, Section 4.3 addresses the latest trends related to generative networks based on patent application dates, while Section 4.4 provides insights into geographical patterns and gaps. 4.1. Main Technology Clusters (RO1) In total, 309 CPC subclasses were identified, covering a wide variety of technological fields. The primary ones were assessed through the SNMF algorithm, which resulted in the extraction of technology clusters. Within the range from 2 to 20 clusters, Fig. 2 presents the variance explained by the examined models along with the topic divergence for the top 3 and top 5 CPC subclasses per cluster. The outcomes suggest that topic divergence is maximized for 10 factors and decreases afterwards, while the variance explained by the models does not improve significantly beyond 4 factors. For these reasons, the following investigations were conducted for 10 factors representing 10 technology clusters. Table 1 showcases the most significant CPC subclasses per cluster (matrix \(\:W\) ), the overall prevalence of each cluster in the patent dataset (matrix \(\:H\) ), and a representative title assigned based on the fields covered by their top CPC classifications. Also, by examining the lower hierarchy levels of the CPC scheme, some more specific concepts were detected based on prevalent CPC groups. Table 1 Main patent technology clusters related to generative networks Cluster Top CPC Subclasses Sum Titles Major concepts 1 G06T (739.12) ; G16H (9.29) ; G01N (6.17) 36.55 Image Data Processing or Generation Image Analysis (G06T7) ; Image enhancement or restoration (G06T5) 2 B60W (591.05) ; G08G (69.37) ; G01S (36.87) ; G05D (34.33) ; G01C (26.27) 1.69 Vehicle and Traffic Control Systems Vehicle control systems (B60W60) 3 G10L (625.04) ; G10H (25.58) ; H04M (11.35) 4.23 Speech Analysis Speech Recognition (G06L15) ; Speech synthesis; Text to speech systems (G06L13) 4 G06N (752.27) ; G16H (13.68) ; G16C (7.03) 111.93 Computing Arrangements Based on Specific Computational Models Neural Networks (G06N3/02) ; Machine learning (G06N20) 5 H04N (645.12) ; H03M (19.99) ; G09G (9.88) 3.91 Pictorial Communication Selective content distribution (H04N21) ; Methods or arrangements for processing digital video signals (H04N19) 6 G06Q (650.52) ; H02J (116.62) ; Y04S (18.94) 8.17 Business ICTs Administration; Management (G06Q10) ; Commerce (G06Q30) ; Circuit arrangements for AC mains or AC distribution networks (H02J3) 7 G06V (746.65) ; Y02T (11.48) ; Y02P (4.24) 34.84 Image or Video Recognition or Understanding Image or video recognition or understanding (G06V10) ; Recognition of biometric, human-related or animal-related patterns (G06V40) 8 H04L (646.41) ; H04W (31.22) ; G16Y (16.35) 4.29 Transmission of Digital Information Network security (H04L63) ; Cryptography and security protocols (H04L9) ; Network services or applications (H04L67) ; User-to-user messaging (H04L51) 9 G06F (740.47) ; G01R (3.77) ; G01M (3.73) 41.66 Electrical Digital Data Processing Pattern recognition (G06F18) ; Information retrieval; Database structures therefor; File system structures therefor (G06F16) ; Handling natural language data (G06F40) ; Computer-aided design [CAD] (G06F30) ; Protection from unauthorised activity (G06F21) 10 A61B (646.8) ; G16H (223.57) ; A61N (13.26) 5.581 Healthcare-related Diagnosis, Surgery and Identification ; Healthcare informatics Measuring for diagnostic purposes (A61B5) ; Apparatus or devices for radiation diagnosis and radiation therapy equipment (A61B6) ; Diagnosis using ultrasonic, sonic or infrasonic waves (A61B8) ; ICT for medical diagnosis, simulation data mining, epidemics or pandemics (G16H50) ; Handling or processing of medical images (G16H30) The definitions of the representative CPC subclasses indicate that generative networks are primarily combined with other computational models (Cluster 4), mostly neural networks and machine learning technologies, to offer utilities and address various tasks. Beyond technologies of this type, the remaining major fields concern data processing (Cluster 9) and image analysis (Clusters 1 and 7). It should be noted that the main difference between the subclasses G06T (Cluster 1) and G06V (Cluster 7) is that the former one concerns data processing and transformations. At the same time, the latter one is closer to pattern recognition and machine learning. The rest of the clusters suggest the involvement of generative networks in subjects that are not directly related to AI as well. These clusters concern healthcare (Cluster 10), vehicle and traffic control systems (Cluster 2), and business ICTs (Cluster 6). In contrast to the aforementioned areas of interest, Clusters 5 and 8 can be seen as types of methods rather than applications, meaning that generative networks are combined with data transmission systems (Cluster 8) and pictorial communication technologies (Cluster 5) in various services. Lastly, while Cluster 3 highlights the significance of speech analysis with the use of generative networks, its relatively small prevalence demonstrates the interest of organizations towards image analysis rather than audio-music data. In total, these outcomes confirm the relevance of generative AI to image and audio data as well as its adoption in diverse applications, showing that organizations have already turned their attention to technologies of this type to improve their services and internal procedures. 4.2. Technology convergence (RO1) The previous section showed that there are at least 10 distinct technological directions associated with generative networks, as expressed by CPC subclasses. However, more specific technological associations can be described through pairs of CPC subclasses with strong convergence. Figure 3 presents the convergence network between the observations that occurred in at least 20 patent families. The subclass G06N was excluded from the analysis since it occurred in more than 20000 families. The size of each node is analogous to the number of patent families that each subclass occurred in, i.e., larger nodes correspond to more frequent observations. The outcomes of the proposed approach suggest that there are three main nodes representing multiple convergence opportunities that can be studied for future projects and innovations. Undoubtedly, digital data processing methods (G06F) constitute the most necessary subclass to cover a variety of tasks related to generative networks. Despite the large prevalence of the CPC subclasses G06T and G06V, the CPC subclass G06F had a notably higher indegree, i.e., 42 against 15 and 10, respectively. Additionally, there are some smaller communities formed by the representative CPC subclasses of the major technology clusters (Section 4.1), more particularly G16H and A61B, H04L, G10L, and G06Q. Thus, through the above network, it is evident that while the investigated patent families can be grouped into 10 large technology clusters, more specific use cases also exist. In total, 86 connections (edges) were created, highlighting the existence of technology convergence concepts and potential opportunities covering a wide range of technological fields. Beyond the primary structure of the convergence network, it is worth highlighting some of the major convergence observations that reflect both high II evaluations and co-occurrence frequencies. Starting from G06F, the most notable opportunities concern the inclusion of digital data methods in business ICTs (G06Q), systems concerning data transmission mechanisms (H04L), speech analysis (G06F), measuring electric variables (G01R), and communication networks (H04W). Other major convergence opportunities are related to applications regarding image processing-generation (G06T) and image recognition techniques (G06V); image processing-generation (G06T) and pictorial communication (H04N); image processing-generation (G06T) and healthcare-related technologies (A61B); image processing-generation (G06T) and material analysis (G01N). These insights indeed confirm multiple pathways that can inspire and guide any researcher in future projects, as proposed by the industrial and academic activity represented by patent data. As a use case, the above network can also be studied according to one or multiple target CPC subclasses to identify implementations that can elevate the potential of a complex technology or include necessary types of systems and methods represented by prevalent CPC subclasses. 4.3. Trend Analysis (RO2) To recognize temporal trends and the evolution of generative networks, it is worth evaluating the different states of associated technological directions over time. Below, the annual frequencies of new patented inventions related to generative networks are presented (Fig. 4 ), where it should be noted that only 44 new patent families were disclosed prior to 2016. The latest application date in the dataset was in January 2026, while the most recent earliest filing date among patent families was in December 2025. The analysis proves that there is indeed a significant amount of new patent families after 2022, probably because of the disclosure of new generative AI tools offering advanced capabilities and use cases. Furthermore, Fig. 5 demonstrates the annual patent share of the 10 technology clusters, showing yearly changes and the most prolific years per technology. To provide a comparable view, the evaluations regarding yearly patent share were scaled to sum to 1 for each cluster using min-max scaling. For similar reasons, only the most significant CPC subclasses per cluster were displayed, as they cover the majority of each cluster’s information. The evaluations indicate the emergence and decline of distinct technology clusters during the investigated time period, as both positive and negative trajectories are observed. Although neural networks and machine learning (G06N) and speech analysis techniques (G10L) peaked in 2019, they are characterized by a declining trajectory till 2023, when their annual shares stabilize. Similarly, technologies concerning image analysis and generation (G06T) were significantly developed after 2017, but also experienced similar progress after 2021. In contrast, while both digital data processing (G06F) and image pattern recognition (G06V) were less significant till 2022, they have been in demand ever since. Especially, the overall frequency of data processing technologies increased by more than 100% from 2022 to 2025. Moving towards minor concepts, the evaluations suggest the recent rise of generative networks in business technologies, indicating their potential in tasks regarding administration and management (G06Q10) and commerce (G06Q30). The remaining technology clusters reached their highest patent share before 2022, highlighting the decline of related fields and the emergence of others. According to patent share, technologies related to data processing (G06F), image pattern recognition (G06V), and business tasks (G06Q) can be characterized as the only trending patent topics, since they are the only technology clusters that show a significant rise after 2022. The yearly technology share was further used to gain a deeper understanding of the growth of each technology cluster with respect to its overall prevalence instead of the yearly patent frequency (Fig. 6 ). In contrast to the previous evaluations, technology share suggests that the overall prevalence of most technology clusters grew significantly after 2022, but some emerged more rapidly. This is the main reason behind the consistent decrease in patent share of many technology clusters (Fig. 5 ). Nonetheless, technologies related to healthcare (A61B) and vehicles (B60W) exhibited the most significant decline, as their patent share peaked in 2020 and decreased thereafter. In summary, the analysis suggests that while there were distinct technology directions that caught the attention of organizations and practitioners over time, the use and orientation of generative networks did not shift from fundamental concepts, e.g., neural networks, machine learning, and image and speech analysis. Transitioning to more specific subjects covered by individual CPC subclasses, Table 2 provides the technologies characterized with the highest and lowest self-growth evaluations, along with their total frequency in the entire dataset. Starting from trending technologies, the outcomes suggest that patents related to generative networks have recently included functionalities that consider energy efficiency, climate change mitigation, and pollution (Y02T, Y02D, Y02A, Y02P, and Y04S). The current analysis confirms and highlights the significance and necessity of these domains in modern implementations, which are generally regarded as mandatory for a green transition. Another arising subject is the use of generative networks in technologies concerning measuring and testing purposes (G01D, G01M, G01J, G01W, and G01H). Lastly, the remaining CPC technologies relate to more concrete subjects, indicating the recent emergence of IoT (G16Y), graphical data reading (G06K), and power generation applications with the use of generative AI. Table 2 Trending and declining types of technologies Most Trending Technologies Most Declining Technologies CPC Self-growth Frequency CPC Self-growth Frequency Y02T 33.659 1525 G03F 0.281 146 Y02D 24.357 355 G05D 0.283 254 Y02A 16.226 534 A61N 0.288 143 Y02P 15.677 517 H10P 0.325 106 G01D 11.889 116 B60R 0.343 47 G16Y 11.364 136 H04R 0.463 60 G01M 10.227 247 A61C 0.533 69 Y04S 10.000 308 A63B 0.609 103 Y02E 9.375 83 A63F 0.620 371 G06K 9.125 81 G10H 0.634 379 G01J 7.571 60 G11B 0.652 76 H02S 7.000 32 G05B 0.669 489 Y10S 4.200 26 B60L 0.744 136 G01W 3.000 100 H03M 0.756 72 G01H 2.750 45 E21B 0.789 68 On the other end, the evaluations regarding technologies that gained less attention than others in recent years further prove the existence of declining themes as well, since the lowest observed self-growth is close to 0.28 while the highest is close to 34. In this case, the CPC subclasses could not be grouped according to similar encodings and characteristics. This means that many diverse topics constituted temporal trends that potentially did not align with the utilities of generative networks, thus not experiencing significant progress when other types of technologies were brought to the surface. In summary, the extremely high evaluations show the consistent transition of technologies into new horizons, ideas, and opportunities that should be monitored by researchers intending to develop innovative solutions. Given the findings of the analysis, these solutions should not be directed to outdated subjects but to modern methods while taking into account existing principles like climate mitigation. 4.4. Geographical Overview (RO3) Through a different perspective, the current section offers information about country-wise technological patterns. Jurisdictions that occurred in 200 (close to 1%) and more patent families were investigated to address only meaningful patterns and not random events. It should be mentioned that only two of the remaining concurrent jurisdictions were observed in over 50 patent families, and another three over 20. Below, Table 3 provides the coverage of each jurisdiction per cluster, along with the total frequency of each jurisdiction. Table 3 Jurisdiction shares per technology cluster Jurisdiction Frequency G06T B60W G10L G06N H04N G06Q G06V H04L G06F A61B China 0.721 0.777 0.549 0.591 0.667 0.628 0.639 0.850 0.606 0.711 0.580 USA 0.279 0.353 0.765 0.571 0.426 0.506 0.249 0.231 0.430 0.286 0.524 Korea 0.101 0.123 0.246 0.209 0.095 0.276 0.196 0.072 0.112 0.134 0.200 Europe 0.100 0.133 0.312 0.210 0.162 0.287 0.096 0.089 0.200 0.101 0.254 Japan 0.054 0.079 0.140 0.138 0.085 0.178 0.046 0.053 0.062 0.052 0.164 Germany 0.024 0.039 0.201 0.037 0.038 0.033 0.005 0.027 0.014 0.019 0.029 Canada 0.013 0.018 0.012 0.023 0.024 0.023 0.030 0.012 0.024 0.015 0.079 Australia 0.012 0.021 0.031 0.023 0.019 0.030 0.029 0.013 0.025 0.015 0.061 Great Britain 0.011 0.023 0.033 0.016 0.019 0.041 0.005 0.010 0.019 0.009 0.026 Taiwan 0.010 0.021 0.001 0.015 0.015 0.034 0.005 0.009 0.024 0.010 0.012 The outcomes show the existence of distinct countries with high involvement, marking their significant technological status and overall development compared to other countries. Undoubtedly, most innovations are protected in China, which was related to more than 70% of the patent families, while the second most prevalent country (USA) covered less than 30%. The high coverage of Chinese applications affects the significance of other countries since they dominate in all major technology clusters. For each jurisdiction, neural networks and machine learning classifications (G06N) cover a large proportion of patent classifications except for Korean patents, where technologies of this type are not as frequent. Moreover, Chinese patents include fewer CPC subclasses related to this cluster on average. Overall, the relative frequencies indicate that investors operating in China are more focused on image data (G06V and G06T) and data processing methods (G06F) than other areas of interest, showing a clear direction toward providing image analysis tools. In fact, technologies associated with speech analysis (G10L), vehicles (B60W), and healthcare (A61B) were the only ones where USA applications had coverage, as Chinese applications did. When considering the total frequency of jurisdictions, inventions related to vehicles (B60W) are notably more frequent in Germany than in any other jurisdiction, while a large proportion of the related patents are protected in the USA. Regarding speech analysis (G10L), the USA, Korea, Europe, and Japan patents constitute the most relevant to technologies of this type. Similarly, healthcare-related inventions (A61B) and technologies related to businesses (G06Q) were more developed in Canada and Australia than expected, while Korean applications included the latter types of technologies more frequently than other countries as well. Other notable evaluations concern Taiwan and Great Britain, which are closer to image processing and generation (G06T) and pictorial communication (H04N) technologies, while Taiwan is closer to network systems (H04L) than the rest of the jurisdictions as well. In summary, the primary item of interest was to demystify whether there are economic opportunities and available resources that are driven by country systems that affect the investments and involvement of third parties interested in generative networks. While the outcomes show that most inventions are protected in China, the orientation of the organizations operating in the various countries is quite different. This finding suggests that the culture, background, resources, and economic system of a country can affect domestic activities, reflecting the attraction and establishment of significant organizations, investors, and individual concepts. When it comes to practical contributions, the current section offered information about leading applicants from a geographical perspective, where it was proved that China dominates the global market and the USA is the second most significant country. The landscape of the different countries indicated the connection between geographical characteristics and technologies surrounding generative networks. These characteristics can drive opportunities with high potential at the national level because of a suitable geographical position, primary sector, high-end technological systems, education systems, traditions, and workforce. Finally, evaluating the expertise and gaps within countries can guide future researchers in gathering information about the existing market and designing appropriate business plans. 5. Conclusion Generative AI changed the landscape of industrial innovations and global investments by offering real-time solutions for generating useful content according to user needs and purposes. Through a patent analysis reviewing industrial developments and technological patterns, the current study offered insights into technological patterns reflecting the potential of generative networks and their overall development across time. The analysis was focused on the CPC classifications used to describe the characteristics of patents, CPC subclasses to be more precise, setting the basis towards addressing major technological directions and convergence opportunities (RO1), recent trends (RO2), and geographical patterns (RO3). A total of 309 technologies were summarized into 10 clusters covering image and audio data, data processing methods, business tasks, healthcare-related implementations, vehicle systems, as well as transmission and communication technologies. These types of technologies were highlighted as the primary use cases and methods associated with generative networks according to patent data. By identifying strong associations between individual types of technologies, the paper also demonstrated a convergence network including 86 convergence opportunities that were previously observed in patent data related to generative networks. Furthermore, a later trend analysis approach was followed to evaluate the progress of the various to classify trending and declining concepts since technologies evolve. This approach indicated pathways for providing up-to-date services and products and responding to the market needs as indicated by existing patented technologies. Fields related to data processing methods, business ICTs, and image pattern recognition were classified as the most trending major concepts. A complementary evaluation process also led to discovering the recent significance of power efficiency, climate change, and measuring-testing techniques, thus establishing new technological requirements for efficient and green technologies. In the last part of the paper, the main item of interest was to clarify any potential geographical patterns that would give clues about the technological status and economic activity of distinct countries. More than 70% of the innovations were covered in China, while the USA was the second most significant country involved in almost 30% of the related patent families. Despite the low frequency of other countries, the investigated patent data led to the identification of the diverging technological orientations across countries, pinpointing the existing developments of vehicle control systems in Germany, the high frequency of technologies related to image data in China, the diversity of Korean applications, and other significant findings. All in all, the current study can offer a basis for technological patterns related to generative networks from different perspectives. Essential concepts and convergence opportunities were discovered and discussed through cluster analysis and association rules, providing information regarding complex technological interconnections that demonstrate the overall potential and use cases of generative networks. The discovery of the latest trends is valuable knowledge for adapting and empowering projects with new ideas, utilities, and improvements with up-to-date technologies. Geographical patterns can be considered a reflection of the industrial and academic activities within a country, which concern available resources, technological status, and investment prospects. When it comes to future work, the current study was dedicated to all types of technologies associated with generative AI. However, individual use cases like business ICTs can be further investigated to complement the findings provided in the current study. Furthermore, patent data includes information about the main inventors and applicants related to each patent. By analyzing this type of information, future projects can develop approaches to develop collaboration networks or address the task of competitor analysis by comparing patent portfolios. Lastly, comparing patent data with other resources, including academic literature and industrial data, is also a noteworthy idea that can reveal gaps and opportunities between sectors. Declarations Funding: No funding was received for conducting this study. Author Contribution C.K. authored the main manuscript, performed data collection and analysis, and designed the figures and table layouts. Acknowledgments: Not Applicable. Availability of data and materials: The data can be publicly accessed through the Patent Lens Database. References Sengar, S. S., Hasan, A. B., Kumar, S., & Carroll, F. (2025). Generative artificial intelligence: a systematic review and applications. Multimedia Tools and Applications, 84(21), 23661-23700. Mittal, U., Sai, S., Chamola, V., & Sangwan, D. (2024). A comprehensive review on generative AI for education. Ieee Access, 12, 142733-142759. Shokrollahi, Y., Yarmohammadtoosky, S., Nikahd, M. M., Dong, P., Li, X., & Gu, L. (2023). A comprehensive review of generative AI in healthcare. arXiv preprint arXiv:2310.00795. Shafiee, S. (2025). Generative AI in manufacturing: a literature review of recent applications and future prospects. Procedia CIRP, 132, 1-6. Kar, A. K., Varsha, P. S., & Rajan, S. (2023). Unravelling the impact of generative artificial intelligence (GAI) in industrial applications: A review of scientific and grey literature. Global Journal of Flexible Systems Management, 24(4), 659-689. Fujii, H., & Managi, S. (2018). Trends and priority shifts in artificial intelligence technology invention: A global patent analysis. Economic Analysis and Policy, 58, 60-69. Abadi, H. H. N., & Pecht, M. (2020). Artificial intelligence trends based on the patents granted by the United States patent and trademark office. IEEE Access, 8, 81633-81643. Liu, N., Shapira, P., Yue, X., & Guan, J. (2021). Mapping technological innovation dynamics in artificial intelligence domains: Evidence from a global patent analysis. Plos one, 16(12), e0262050. Chang, S. H. (2021). Technical trends of artificial intelligence in standard-essential patents. Data Technologies and Applications, 55(1), 97-117. Giczy, A. V., Pairolero, N. A., & Toole, A. A. (2022). Identifying artificial intelligence (AI) invention: A novel AI patent dataset. The Journal of Technology Transfer, 47(2), 476-505. Ailia, M. J., Thakur, N., Abdul-Ghafar, J., Jung, C. K., Yim, K., & Chong, Y. (2022). Current trend of artificial intelligence patents in digital pathology: a systematic evaluation of the patent landscape. Cancers, 14(10), 2400. Lee, M. (2020). An analysis of the effects of artificial intelligence on electric vehicle technology innovation using patent data. World Patent Information, 63, 102002. Benjamens, S., Dhunnoo, P., Görög, M., & Mesko, B. (2023). Forecasting artificial intelligence trends in health care: systematic international patent analysis. Jmir Ai, 2(1), e47283. Abbas, A., Zhang, L., & Khan, S. U. (2014). A literature review on the state-of-the-art in patent analysis. World Patent Information, 37, 3-13. Pejić Bach, M., Pivar, J., & Fabac, R. (2024). Emerging trends in generative artificial intelligence: Insights from patent analysis using Lens. org toolkit. ITIS 2024 BOOK OF PROCEEDINGS''Harnessing the power of big data for green and digital transition'', 151-160. Charmanas, K., Mittas, N., & Angelis, L. (2023a). Topic and influence analysis on technological patents related to security vulnerabilities. Computers & Security, 128, 103128. Charmanas, K., Georgiou, K., Mittas, N., & Angelis, L. (2023b). Classifying the main technology clusters and assignees of home automation networks using patent classifications. Computers, 12(10), 211. Yang, G. C., Li, G., Li, C. Y., Zhao, Y. H., Zhang, J., Liu, T., ... & Huang, M. H. (2015). Using the comprehensive patent citation network (CPC) to evaluate patent value. Scientometrics, 105(3), 1319-1346. Choi, D., & Song, B. (2018). Exploring technological trends in logistics: Topic modeling-based patent analysis. Sustainability, 10(8), 2810. Choi, H., Oh, S., Choi, S., & Yoon, J. (2018). Innovation topic analysis of technology: The case of augmented reality patents. IEEE Access, 6, 16119-16137. Hu, X., Rousseau, R., & Chen, J. (2012). A new approach for measuring the value of patents based on structural indicators for ego patent citation networks. Journal of the American Society for Information Science and Technology, 63(9), 1834-1842. Lee, M., & Lee, S. (2017). Identifying new business opportunities from competitor intelligence: An integrated use of patent and trademark databases. Technological Forecasting and Social Change, 119, 170-183. Haghighian Roudsari, A., Afshar, J., Lee, W., & Lee, S. (2022). PatentNet: multi-label classification of patent documents using deep learning based language understanding. Scientometrics, 127(1), 207-231. Trappey, A. J., Trappey, C. V., Govindarajan, U. H., & Sun, J. J. (2019). Patent value analysis using deep learning models—The case of IoT technology mining for the manufacturing industry. IEEE Transactions on Engineering Management, 68(5), 1334-1346. Karataş, A. R., Kazak, H., Akcan, A. T., Akkaş, E., & Arık, M. (2024). A bibliometric mapping analysis of the literature on patent analysis. World Patent Information, 77, 102266. Lin, T. Y., & Chou, L. C. (2025). A systematic review of artificial intelligence applications and methodological advances in patent analysis. World Patent Information, 82, 102383. Shomee, H. H., Wang, Z., Ravi, S. N., & Medya, S. (2025, July). A survey on patent analysis: From nlp to multimodal ai. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) (pp. 8545-8561). Verendel, V. (2023). Tracking artificial intelligence in climate inventions with patent data. Nature Climate Change, 13(1), 40-47. Santarelli, E., Staccioli, J., & Vivarelli, M. (2023). Automation and related technologies: a mapping of the new knowledge base. The Journal of Technology Transfer, 48(2), 779-813. Hu, T., Guo, J., Zhang, T., Liu, J., Sun, X., & Chang, Z. (2022, June). Analysis of the development trend of artificial intelligence technology application in the field of sports-based on patent measurement. In International Conference on Human-Computer Interaction (pp. 337-354). Cham: Springer International Publishing. Park, H. S. (2017). Technology convergence, open innovation, and dynamic economy. Journal of Open Innovation: Technology, Market, and Complexity, 3(4), 1-13. Charmanas, K., Filippou, K., Mittas, N., & Angelis, L. (2025a). Exploring the role of information security news descriptions on retweet proneness and user interactions. Journal of Computational Social Science, 8(3), 60. Lee, D. D., & Seung, H. S. (1999). Learning the parts of objects by non-negative matrix factorization. nature, 401(6755), 788-791. Lee, D., & Seung, H. S. (2000). Algorithms for non-negative matrix factorization. Advances in neural information processing systems, 13. Kim, H., & Park, H. (2007). Sparse non-negative matrix factorizations via alternating non-negativity-constrained least squares for microarray data analysis. Bioinformatics, 23(12), 1495-1502. Žitnik, M., & Zupan, B. (2012). Nimfa: A python library for nonnegative matrix factorization. The Journal of Machine Learning Research, 13(1), 849-853. Dieng, A. B., Ruiz, F. J., & Blei, D. M. (2020). Topic modeling in embedding spaces. Transactions of the Association for Computational Linguistics, 8, 439-453. Zhang, S., & Wu, X. (2011). Fundamentals of association rules in data mining and knowledge discovery. Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, 1(2), 97-116. He, Q. (1999). Knowledge discovery through co-word analysis. Charmanas, K., Georgiou, K., Papageorgiadis, K., Mittas, N., & Angelis, L. (2025b). A topic-oriented trend analysis framework for Stack Exchange questions: Case study on ChatGPT related queries on Stack Overflow. Information and Software Technology, 107969. Footnotes https://www.uspto.gov/web/patents/classification/cpc/html/cpc.html https://www.lens.org/ https://www.epo.org/en 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-9446615","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":624822393,"identity":"7d870619-1e58-47fd-8100-cee7d2aa3fb8","order_by":0,"name":"Konstantinos Charmanas","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA5ElEQVRIiWNgGAWjYBACCQYeCIMfzAECA6K1SDZAtEgQr8XgALFaJNvPHvxcmWOXZ3z8+MMbDDWH68wZ2K99wKdFmicvWfLstuRiszM5xhYMxw5LWDbwFM/Ap0VOgsdAsnEbc+K2GzxsEgxshyUMDvAk43UYUIvxz8Zt9YmbZ7A/k2D4R4QWaQkeM6AthxM3SDCYSTC2gbSwH8arRbInx8yycdvxxBkgvyT2pUtuOMzDjFeLxPEzxjcbt1Un9rcDQ+zDN2t+g+Ptj/FqQQUJIIKZh2DUYAD2ByRrGQWjYBSMgmENAAwARXbKiuVKAAAAAElFTkSuQmCC","orcid":"","institution":"Aristotle University of Thessaloniki","correspondingAuthor":true,"prefix":"","firstName":"Konstantinos","middleName":"","lastName":"Charmanas","suffix":""}],"badges":[],"createdAt":"2026-04-17 08:54:44","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-9446615/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9446615/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107696977,"identity":"fa3a8e6e-c7d7-4f91-b510-1f95d4143944","added_by":"auto","created_at":"2026-04-24 07:18:53","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":81906,"visible":true,"origin":"","legend":"\u003cp\u003eOverview of the study\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-9446615/v1/9ebc8c0f9fa6d432e5d82e2c.png"},{"id":107707506,"identity":"3b795aeb-a401-4b82-910f-0b0daf1f8152","added_by":"auto","created_at":"2026-04-24 09:20:28","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":100485,"visible":true,"origin":"","legend":"\u003cp\u003eModel evaluations per rank\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-9446615/v1/ae90b52024419574c2b4e8d4.png"},{"id":107696982,"identity":"0d2cda0b-20ca-4795-a3fa-d98be9200d7c","added_by":"auto","created_at":"2026-04-24 07:18:53","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":209459,"visible":true,"origin":"","legend":"\u003cp\u003eTechnology convergence network\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-9446615/v1/0088f7ec55740b3bd9f462c5.png"},{"id":107696981,"identity":"1a44ba0b-b0a1-4049-8313-285d9863b652","added_by":"auto","created_at":"2026-04-24 07:18:53","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":53588,"visible":true,"origin":"","legend":"\u003cp\u003eNew Patent Families Disclosed per Year\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-9446615/v1/f5d7dabf23138dd21d45a9f0.png"},{"id":107696978,"identity":"91f7a491-1c25-4ed1-947a-38cd92d9ff21","added_by":"auto","created_at":"2026-04-24 07:18:53","extension":"jpeg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":176802,"visible":true,"origin":"","legend":"\u003cp\u003eYearly patent share of technology clusters\u003c/p\u003e","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-9446615/v1/047a837c9b16848abb38ce75.jpeg"},{"id":107696979,"identity":"e9e0637a-efb7-4f1b-8f24-342694559092","added_by":"auto","created_at":"2026-04-24 07:18:53","extension":"jpeg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":151606,"visible":true,"origin":"","legend":"\u003cp\u003eYearly technology share of technology clusters\u003c/p\u003e","description":"","filename":"floatimage6.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-9446615/v1/bc69c74fe6aeee98e09f6470.jpeg"},{"id":108006048,"identity":"bc2e4b2a-a927-439f-9387-52bd960010d0","added_by":"auto","created_at":"2026-04-28 12:52:27","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1090728,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9446615/v1/2c05d208-db86-4ff2-8aa1-f9d658ec35eb.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Patent Landscape of Generative Networks: A Data-Driven Examination of Essential Concepts Convergence Opportunities and Technological Trends Through Patent Analysis","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eTechnological patents reflect the exclusive rights granted for valuable inventions, whose characteristics and patterns are protected for competitive and strategic reasons. The potential use, reproduction, or any kind of exploitation of an invention by third parties is successfully restricted so that the owners are the only ones benefiting from their offerings. Patents signify the importance of distinct inventions, which can cover breakthroughs improving the value of products, systems, and services or even the evaluation of a whole organization.\u003c/p\u003e\n\u003cp\u003eOne major technological field characterized by a massive rise and individual breakthroughs is generative Artificial Intelligence (AI), which primarily covers technologies for image, video, and natural language processing or generation [1]. These offerings highlighted the emergence of technologies in various areas of interest, including education [2], healthcare [3], and manufacturing [4]. While there is considerable research through scientific [1] and grey literature [5] reviews, patent data constitute a less popular source for overviewing industrial and research activities.\u003c/p\u003e\n\u003cp\u003eNonetheless, researchers have already shifted their attention towards patents related to AI, covering AI in general [6-10] and specific subjects, such as digital pathology [11], electric vehicles [12], and healthcare [13]. The broader term used to label studies dedicated to examining and summarizing patent data is called patent analysis, where some common objectives include trend and competitor analysis, technology forecasting, and assessing patent quality [14]. In the case of generative AI, research from Bach et al. [15] constitutes a preliminary patent analysis focusing on the Cooperative Patent Classification (CPC\u003ca href=\"#_ftn1\" name=\"_ftnref1\" title=\"\"\u003e\u003c/a\u003e\u003csup\u003e1\u003c/sup\u003e) scheme and the technical content of patent documents. Their goal was to search for technologies co-occurring with generative AI, hence providing valuable insights into prevalent subjects.\u003c/p\u003e\n\u003cp\u003eBy considering the rise of generative AI and valuable information included in patent data, the current study builds upon the outcomes produced by previous research by investigating technology-based patterns related to generative networks through global patent analysis. The primary Research Objectives (ROs) were the following:\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRO1\u003c/strong\u003e: Address major technological directions and convergence opportunities\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRO2\u003c/strong\u003e: Find emerging and declining technological concepts\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRO3\u003c/strong\u003e: Evaluate geographical patterns and differences\u003c/p\u003e\n\u003cp\u003eTo cover these objectives, occurrence and co-occurrence statistics, cluster analysis, and association rules constituted the main methods employed, while the Patent Lens Database\u003ca href=\"#_ftn2\" name=\"_ftnref2\" title=\"\"\u003e\u003c/a\u003e\u003csup\u003e2\u003c/sup\u003e and the CPC scheme were the primary sources of information. In fact, patent CPC classifications were investigated as the ground truth to explore technology frequencies and patterns. To be more precise, RO1 was addressed through cluster analysis and association rules, where strong associations led to the development of a convergence network. In combination with cluster properties and CPC classifications, major trends were evaluated through patent application dates, while geographical patterns were assessed based on patent jurisdictions.\u003c/p\u003e\n\u003cp\u003eIn total, this study fills an existing gap regarding the patent activity surrounding generative networks via the provision of information regarding major and trending technologies and the insights concerning geographical and technological convergence. This information can contribute to future projects by offering guidelines about essential technologies considered by large companies and about more specific opportunities, adapting to the needs of a potential researcher, as reflected by geographical opportunities and convergence options of distinct applications of interest with other technological domains.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe following section discusses the contribution and themes covered by previous patent analysis studies, focusing especially on papers dedicated to patents related to AI (Section 2). Section 3 presents the methodology employed to cover the ROs, while the following section includes the main outcomes of the analysis (Section 4). The paper concludes with Section 5 by discussing concluding remarks and recommendations for future research.\u0026nbsp;\u003c/p\u003e"},{"header":"2. Related Work","content":"\u003cp\u003ePatent analysis is the review of patented inventions covering any technological field and related contributors that shape the existing knowledge background and status of the field. Abbas et al. [14] describe a generic workflow for patent analysis and suggest that patent documents contain valuable information for research purposes, including technical characteristics (title, claims, abstract, description, patent classifications) and associated metadata such as patent inventors, applicants, and citations. After establishing structured patent data, researchers can evaluate data patterns to cover various objectives and challenges that could offer valuable insights about the current and future status of a technological field.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn our previous research [16, 17], we also investigated properties of this type to identify primary and trending-declining topics and major contributors (inventors and applicants), as well as assess patent value through citation networks. The ideas and methodologies were inspired by previous research that set the basis for extracting patterns through appropriate indicators [18, 19], while many other research studies have previously proposed related approaches for capturing semantics [20], analyzing citation networks [21], competitor analysis [22], patent classifications [23], and patent value assessment [24].\u003c/p\u003e\n\u003cp\u003eA later bibliometric analysis of the literature [25] indicates that patent analysis is widely employed to investigate diverging technological areas of interest, as suggested by the journals publishing relevant material. These journals may range from innovation science, R\u0026amp;D, and business topics to computer science, e.g., information processing and expert systems. Based on their analysis, some more domain-specific themes related to engineering and energy were among the most frequent fields as well.\u003c/p\u003e\n\u003cp\u003eFurthermore, AI methods have also been implemented in patent analysis frameworks for multiple relevant tasks. Lin and Chou [26] showed that AI finds implementations in tasks beyond predictive mechanisms (trend prediction and value assessment), including the evaluation of topics, emerging technologies, network graphs, and firm activities. Through a review dedicated to pretrained and large language models, Shomee et al. [27] highlight the value of these state-of-the-art models in patent analysis for the tasks of patent classification, quality-value analysis, and text generation. This information shows that AI can be used to address a plethora of tasks covering standard patent analysis scopes and objectives.\u003c/p\u003e\n\u003cp\u003eSince the current study focuses on patents related to generative networks, it is worth discussing key insights and approaches from previous studies exploring patents associated with the general domain of AI. Fuji and Managi [6] were among the first to discuss patent activity surrounding AI, already identifying the involvement of major companies and the rapid activity surrounding relevant technologies. In the investigated timeline, the analysis showed that the attention was shifting from biological and knowledge-based models to mathematical and other AI technologies.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThrough text mining, Abadi and Pecht [7] later showed that fields related to neural networks, machine learning, and pattern recognition, along with specific algorithms like decision trees and support vector machines, were widely employed to develop technologies with AI capabilities. Through the CPC scheme, they also pointed out that techniques concerning digital data processing and transmission, as well as image and speech analysis, constitute some of the major associated methodologies. In addition, some indirectly related fields like pictorial and telephonic communication, business, and healthcare technologies were also apparent.\u003c/p\u003e\n\u003cp\u003eLater research case studies also support the findings discussed above, confirming the rise of AI technologies whilst investigating other aspects as well, like country-level evaluations [8, 9], technology network analysis [9], and patent classification [10]. Among the latest technological fields, these studies also address the inclusion of natural language processing, hardware, computer vision, and telecommunication technologies in patents associated with AI.\u003c/p\u003e\n\u003cp\u003eTo find the use cases of AI in specific sciences, other studies have also dedicated their research to healthcare-related patents and subjects, showing the coverage of AI technologies in ophthalmology, oncology, radiology, cardiology, and general-purpose surgeries [13]. Similarly, Lee [12] discussed a plethora of different AI methods contributing to the various utilities of electric vehicles, with deep learning architectures, clustering algorithms, and decision trees being among the most frequent ones. In general, there are multiple studies highlighting the value of AI in diverging domains, including climate change [28], automation technologies [29], and sports [30].\u003c/p\u003e\n\u003cp\u003eRegarding generative AI, Bach et al. [15] showed that the patent activity surrounding technologies of this type had a dramatic rise around 2023 and 2024, indicating the tendency of major enterprises to invest in this breakthrough technological field. Through content-based analysis, they also revealed some relevant themes mentioned within patents of this type, like natural language processing, response generation, and training machine learning models. By further exploring CPC patent classifications, they also addressed some prevalent types of technologies concerning generative networks, machine learning, autoencoders, and adversarial learning.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn total, our study builds upon these findings and provides insights into technology convergence and trend analysis while, at the same time, exploring geographical patterns. By investigating CPC classifications, the analysis concentrates on well-established technological fields and not on textual descriptions that may lead to irrelevant and repetitive content or non-interpretable topics. On top of that, both frequent and rare classifications are evaluated, while a cluster analysis method is also employed to provide collective information and cover the entirety of the observed classifications.\u003c/p\u003e"},{"header":"3. Methodology","content":"\u003cp\u003eThe current section demonstrates the methodology that led to the extraction of valuable insights from patent data related to generative networks. The first part of the methodology is the process of patent collection and preprocessing (Section 3.1), while the rest of the sections were organized based on the introduced ROs. Accordingly, Section 3.2 presents the methods followed to establish technology clusters and assess technology convergence opportunities (RO1). A trend analysis pipeline is covered in Section 3.3 (RO2), while Section 3.4 describes the statistical approach used to associate geographical information with technologies and detect patterns (RO3). Figure \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e provides an overview of the study.\u003c/p\u003e\n\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\n \u003ch2\u003e3.1. Patent Collection and Preprocessing\u003c/h2\u003e\n \u003cp\u003eInitially, the Patent Lens Database was selected as the primary source of information, as it gathers global patent data and provides access to multiple resourceful metadata, such as application dates, patent classifications, content information (title, abstract), and jurisdictions. After searching the CPC scheme, the only encoding that is directly associated with generative networks is the CPC subgroup G06N3/0475, described as \u003cem\u003eGenerative networks\u003c/em\u003e. This subgroup falls under the CPC subgroups G06N3/03, entitled \u003cem\u003eNeural networks\u003c/em\u003e, and G06N3/04, entitled \u003cem\u003eArchitecture, e.g., interconnection topology\u003c/em\u003e. In this case, the CPC subgroup G06N3/03 is the superconcept of G06N3/04. In general, the CPC scheme is organized into multiple hierarchy levels where the encoding of generative networks (G06N3/0475) can be disassembled as follows: G (Section), 06 (Class), N (Subclass), 3 (Group), 0475 (Subgroup).\u003c/p\u003e\n \u003cp\u003eThis search strategy led to the collection of 44230 patent applications and granted patents. However, patent documents (documents referring to applications, granted patents, etc.) may have overlaps due to patent renewal procedures, innovation variations and versions, as well as partial or complete technology expansions. For this reason, an approach towards merging observations covering similar content was considered to avoid data replication. Databases track the relationships between patent documents through simple and extended patent families. A simple patent family includes patents covering the same invention, while an extended family includes patents with linked inventions covering similar content.\u003c/p\u003e\n \u003cp\u003eBased on the patent collections formed by extended families, a final dataset of 21048 extended patent families was established by merging related documents based on specific rules. The family sizes of the patents indicated that the inventions covered by the collected patents were relatively new, with a median family size equal to 1 and a third quartile (Q3) equal to 2, while there were some cases where the family size was extremely large (maximum value equal to 668). These observations show that the proposed approach of merging similar inventions was vital towards the validity of this study, as there were large patent families that could affect the outcomes significantly.\u003c/p\u003e\n \u003cp\u003eIn the current study, each family was associated with a CPC subgroup and a patent jurisdiction when at least one of its documents was associated with them. The information about the technologies covering each patent family was gathered based on CPC classifications instead of textual information because the various patent offices use different languages in textual information (titles and abstracts), while the CPC scheme is used globally. When it comes to date information, the earliest application dates of the retrieved documents were stored per family.\u003c/p\u003e\n \u003cp\u003eTo offer a solution that considers both the relationships between CPC encodings (avoid overlaps) and the patterns from different encodings, the CPC subclasses were investigated instead of the CPC subgroups. As a result, each patent family was associated with a CPC subclass with count indicators, according to the total number of CPC subgroups that fall under the subclass and were assigned to the patent family, and binary indicators denoting relevance. It should be noted that the CPC subgroup G06N3/0475 was excluded from the analysis, as it was used to retrieve relevant patents, while 309 CPC subclasses were observed in total.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\n \u003ch2\u003e3.2. Analyzing Technology Co-occurrence Patterns\u003c/h2\u003e\n \u003cp\u003eOverall, the primary focus of the current study is the investigation of patterns surrounding generative networks, as expressed by valuable inventions covered by patents. In this context, technology patterns were investigated through clustering algorithms (Section 3.2.1) and co-occurrence networks (Section 3.2.2). Multivariate analysis was used to cluster together multiple types of technologies based on CPC frequencies, hence finding the general technological directions associated with generative networks. Co-occurrence networks can be seen as a different but complementary approach, evaluating the convergence between pairs of technologies instead of groups, hence establishing convergence networks. Technology convergence is a valuable source of information towards identifying and monitoring emerging opportunities in the market [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e], which in this case are measured based on the total interactions between two different types of technologies in patent data.\u003c/p\u003e\n \u003cdiv id=\"Sec6\" class=\"Section3\"\u003e\n \u003ch2\u003e3.2.1. Technology Clustering\u003c/h2\u003e\n \u003cp\u003eClustering algorithms usually rely on feature (data points) or similarity matrices. Accordingly, a Patent to CPC Matrix (PCM) was created, where each cell denoted the frequency of a CPC subclass (column) in a patent family (row). Similar to our previous research [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e], a variation of Non-negative Matrix Factorization [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e], described as Sparse Non-negative Matrix Factorization (SNMF) in this case [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e], was employed to find technology clusters via the NIMFA Python library [\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e].\u003c/p\u003e\n \u003cp\u003eNMF models decompose an initial matrix of \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:m\\)\u003c/span\u003e\u003c/span\u003e features (rows) and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:n\\)\u003c/span\u003e\u003c/span\u003e observations (columns) described as \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:V\\)\u003c/span\u003e\u003c/span\u003e (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:m\\:\\times\\:n\\)\u003c/span\u003e\u003c/span\u003e), inverted PCM in this case, into two lower-rank matrices reflecting the associations of \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:r\\)\u003c/span\u003e\u003c/span\u003e (rank) latent features (or factors) with the initial features and observations, usually denoted as \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:W\\)\u003c/span\u003e\u003c/span\u003e (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:m\\:\\times\\:r\\)\u003c/span\u003e\u003c/span\u003e) and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:H\\)\u003c/span\u003e\u003c/span\u003e (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:r\\:\\times\\:n\\)\u003c/span\u003e\u003c/span\u003e), respectively. The general objective of NMF models is to approximate the matrix \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:V\\)\u003c/span\u003e\u003c/span\u003e as the product of \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:W\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:H\\)\u003c/span\u003e\u003c/span\u003e (1).\u003c/p\u003e\n \u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\n \u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ1\" name=\"EquationSource\"\u003e$$\\:{V}^{m\\times\\:n}=\\:{W}^{m\\times\\:r}{H}^{r\\times\\:n}$$\u003c/div\u003e\n \u003cdiv class=\"EquationNumber\"\u003e1\u003c/div\u003e\n \u003c/div\u003e\n \u003cp\u003eThe benefit of SNMF compared to other variations is that it enforces sparse factors, meaning that each latent factor is expected to rely only on a few initial features, leading to more interpretable and distinct factors (technology clusters). In contrast to univariate analysis, this approach facilitates the collective investigation and summarization of all 309 CPC subclasses through data structures that are less complex than PCM, hence enabling pattern recognition and the inclusion of all information about patent classifications.\u003c/p\u003e\n \u003cp\u003eA vital part of the methodology was the selection of an appropriate rank (number of factors), where the approximated low-dimensional matrices (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:W\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:H\\)\u003c/span\u003e\u003c/span\u003e) both capture the variability of the initial matrix (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:V\\)\u003c/span\u003e\u003c/span\u003e) and help summarize data patterns and feature correlations. In this case, the variability explained by each SNMF model and the topic divergence [\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e] between the extracted factors were selected to find the optimal rank, hence leading to factors with divergent compositions that also capture a high proportion of the available information. The first metric is calculated as the percentage of unique top features between all factors, where the top 3 and top 5 features per factor were considered. For example, given a model with two factors and the top 5 features, a topic divergence evaluation equal to 0.9 would mean that only a single feature appeared in the top features of both factors. After finding an appropriate rank, the weights stored in matrix \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:W\\)\u003c/span\u003e\u003c/span\u003e were evaluated to find the technical content covered by each technology cluster, while matrix \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:H\\)\u003c/span\u003e\u003c/span\u003e was considered as the baseline information towards associating technology clusters with jurisdictions and date variables.\u003c/p\u003e\n \u003c/div\u003e\n \u003cdiv id=\"Sec7\" class=\"Section3\"\u003e\n \u003ch2\u003e3.2.2. Constructing Convergence Network\u003c/h2\u003e\n \u003cp\u003eA convergence network is a visual representation of strong associations between two technologies, showing their potential in individual innovations. Strong associations between two observations were evaluated using association rules considering the number of patents including both CPC subclasses at least once as a function of the number of patents including either one. Thus, the current study investigates both CPC frequencies (Section 3.2.1) and binary co-occurrences (Section 3.2.2) to address any potential limitations of each approach.\u003c/p\u003e\n \u003cp\u003eAssociation rule mining is a wide family of data mining methods used to measure the relationships between two items, where one of the essential concepts is described as confidence. Given items \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:X\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:Y\\)\u003c/span\u003e\u003c/span\u003e, the confidence of \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:X-\u0026gt;Y\\)\u003c/span\u003e\u003c/span\u003e measures the probability of \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:Y\\)\u003c/span\u003e\u003c/span\u003e, usually described as consequent, occurring in the observations containing \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:X\\)\u003c/span\u003e\u003c/span\u003e, usually described as antecedent [\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]. This method measures the one-way dependency between the two items, where one is considered the antecedent and the other the consequent. To establish a single indicator, an alternative method described as the Inclusion Index (II) was employed [\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]. II is the maximum probability between \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:X-\u0026gt;Y\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:Y-\u0026gt;X\\)\u003c/span\u003e\u003c/span\u003e,\u003c/p\u003e\n \u003cdiv id=\"Equ2\" class=\"Equation\"\u003e\n \u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ2\" name=\"EquationSource\"\u003e$$\\:{II}_{xy}=\\:\\frac{P\\left(XY\\right)}{min\\left(P\\left(X\\right),P\\left(Y\\right)\\right)}\\:$$\u003c/div\u003e\n \u003cdiv class=\"EquationNumber\"\u003e2\u003c/div\u003e\n \u003c/div\u003e\n \u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\text{P}\\left(\\text{X}\\text{Y}\\right)\\)\u003c/span\u003e\u003c/span\u003e is the probability of \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:X\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:Y\\)\u003c/span\u003e\u003c/span\u003e co-occurring in a dataset while \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\text{P}\\left(\\text{X}\\right)\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\text{P}\\left(\\text{Y}\\right)\\)\u003c/span\u003e\u003c/span\u003e are the probabilities of \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:X\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:Y\\)\u003c/span\u003e\u003c/span\u003e occurring in the dataset, respectively. In the context of the current study, the benefit of II against other options is that this method also addresses the significance of one major technology to other minor fields, thus finding both one-way and two-way convergence opportunities. One-way means that the confidence of either \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:X-\u0026gt;Y\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:Y-\u0026gt;X\\)\u003c/span\u003e\u003c/span\u003e is high, while two-way convergence means that the confidence of both \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:X-\u0026gt;Y\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:Y-\u0026gt;X\\)\u003c/span\u003e\u003c/span\u003e are high.\u003c/p\u003e\n \u003cp\u003eAs a result, high II evaluations can lead to the identification of technology convergence without the need to define antecedents and consequents. For example, generative networks in combination with image analysis techniques may be significant in a variety of technological fields like medical diagnosis, vehicle control systems, pictorial communication, and material examination. In this case, II will respond to high evaluations in individual cases where image analysis techniques are frequent, while two-way approaches would not assess their co-frequencies as significant since these techniques apply to many independent domains. To establish convergence networks, a threshold of 0.4 II evaluations was set to denote strong associations indicating technology convergence between CPC subclasses. Also, only CPC subclasses that occurred in at least 20 patent families (close to 0.1%) were evaluated to avoid random events.\u003c/p\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n \u003ch2\u003e3.3. Finding Technological Trends\u003c/h2\u003e\n \u003cp\u003eDate information can be seen as a ground truth basis towards investigating the evolution and initial disclosure of ideas and implementations associated with a patent dataset to ultimately categorize technology directions as arising, essential, or declining. The earliest application dates of the established patent families were considered to identify arising and declining areas of interest. Starting from clusters, the frequency of each cluster was calculated per year along with two indicators used to evaluate their annual presence, based on the matrix \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:H\\)\u003c/span\u003e\u003c/span\u003e. First, the yearly cluster frequency was calculated. Since the analysis is based on the earliest application dates related to patent families, yearly cluster frequency is related to the disclosure of new inventions per year and the inclusion of technologies covered by each cluster, hence making this information appropriate for evaluating the overall evolution of a technology cluster annually.\u003c/p\u003e\n \u003cp\u003eThe first trend indicator measures the relative frequency of each cluster per patent and year, denoted as yearly patent share. The second indicator measures the relative frequency of a field per year with respect to the total frequency of that field in the investigated patent dataset, denoted as yearly technology share. These two indicators were used to identify the breakthrough years and the overall development of the various technologies over time to categorize them as arising or declining.\u003c/p\u003e\n \u003cp\u003eGoing beyond clusters, individual CPC subclasses that occurred in at least 20 patent families were also examined to provide a more thorough view of emerging technologies. To evaluate trending technologies, the frequencies of different CPC subclasses before and after 2023 were investigated and compared. The selected break point was marked by the launch of ChatGPT and other generative AI tools in late 2022, which undoubtedly changed the landscape of organizational procedures and everyday practices by offering new capabilities to end-users. An approach proposed by Charmanas et al. [\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e] was used to measure self-growth, where the number of topic assignments was replaced with total frequencies of CPC subclasses within a timeframe. In this case, the approach measures each technology\u0026rsquo;s growth after 2023 with respect to its overall frequency prior to this year. A self-growth equal to 2 would mean that the total frequency of the examined CPC subclass increased by 200% after 2023.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\n \u003ch2\u003e3.4. Exploring Geographical Patterns\u003c/h2\u003e\n \u003cp\u003eIndividuals and organizations usually operate in distinct countries and regions rather than globally, meaning that they seek protection for their inventions through patents in the appropriate patent offices. Patent data offer information about the jurisdictions of each patent, which can be used to evaluate regional patterns. This information can be used to address the technological status and differences between various countries. Indicators of this type correspond to enterprise and academic activities that can be reviewed for standard patent analysis scopes like competitor analysis and strategic planning.\u003c/p\u003e\n \u003cp\u003eThis study investigates geographical patterns through technology clusters and major jurisdictions recorded for each patent. Each patent family was associated with a jurisdiction when at least one related patent application was filed to cover the invention in the corresponding country or group of countries, i.e., the European Patent Office (EPO\u003ca href=\"#_ftn1\" name=\"_ftnref1\" title=\"\"\u003e\u003c/a\u003e\u003csup\u003e3\u003c/sup\u003e). Then, the co-frequencies between each jurisdiction and cluster were calculated to compare the various countries according to technology patterns as well as find the most relevant to each cluster. The co-frequencies were standardized based on the total frequency of each cluster, providing ratios instead of absolute frequencies. An evaluation equal to 0.3 would mean that 30% of the investigated cluster is associated with the investigated jurisdiction.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"4. Results and Discussion","content":"\u003cp\u003eSection 4 includes the main findings of this study, addressing the ROs introduced in Section 1 with the use of the methods presented previously (Section 3). Sections 4.1 and 4.2 cover the first RO by investigating technology clusters and technology convergence opportunities, respectively. Furthermore, Section 4.3 addresses the latest trends related to generative networks based on patent application dates, while Section 4.4 provides insights into geographical patterns and gaps.\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e4.1. Main Technology Clusters (RO1)\u003c/h2\u003e \u003cp\u003eIn total, 309 CPC subclasses were identified, covering a wide variety of technological fields. The primary ones were assessed through the SNMF algorithm, which resulted in the extraction of technology clusters. Within the range from 2 to 20 clusters, Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e presents the variance explained by the examined models along with the topic divergence for the top 3 and top 5 CPC subclasses per cluster.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe outcomes suggest that topic divergence is maximized for 10 factors and decreases afterwards, while the variance explained by the models does not improve significantly beyond 4 factors. For these reasons, the following investigations were conducted for 10 factors representing 10 technology clusters. Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e showcases the most significant CPC subclasses per cluster (matrix \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:W\\)\u003c/span\u003e\u003c/span\u003e), the overall prevalence of each cluster in the patent dataset (matrix \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:H\\)\u003c/span\u003e\u003c/span\u003e), and a representative title assigned based on the fields covered by their top CPC classifications. Also, by examining the lower hierarchy levels of the CPC scheme, some more specific concepts were detected based on prevalent CPC groups.\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\u003eMain patent technology clusters related to generative networks\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\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=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\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\u003eTop CPC Subclasses\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSum\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTitles\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMajor concepts\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\u003eG06T (739.12) ; G16H (9.29) ; G01N (6.17)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e36.55\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eImage Data Processing or Generation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eImage Analysis (G06T7) ; Image enhancement or restoration (G06T5)\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\u003eB60W (591.05) ; G08G (69.37) ; G01S (36.87) ; G05D (34.33) ; G01C (26.27)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1.69\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eVehicle and Traffic Control Systems\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eVehicle control systems (B60W60)\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\u003eG10L (625.04) ; G10H (25.58) ; H04M (11.35)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e4.23\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSpeech Analysis\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eSpeech Recognition (G06L15) ; Speech synthesis; Text to speech systems (G06L13)\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\u003eG06N (752.27) ; G16H (13.68) ; G16C (7.03)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e111.93\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eComputing Arrangements Based on Specific Computational Models\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eNeural Networks (G06N3/02) ; Machine learning (G06N20)\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\u003eH04N (645.12) ; H03M (19.99) ; G09G (9.88)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3.91\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePictorial Communication\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eSelective content distribution (H04N21) ; Methods or arrangements for processing digital video signals (H04N19)\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\u003eG06Q (650.52) ; H02J (116.62) ; Y04S (18.94)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e8.17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eBusiness ICTs\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eAdministration; Management (G06Q10) ; Commerce (G06Q30) ; Circuit arrangements for AC mains or AC distribution networks (H02J3)\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\u003eG06V (746.65) ; Y02T (11.48) ; Y02P (4.24)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e34.84\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eImage or Video Recognition or Understanding\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eImage or video recognition or understanding (G06V10) ; Recognition of biometric, human-related or animal-related patterns (G06V40)\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\u003eH04L (646.41) ; H04W (31.22) ; G16Y (16.35)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e4.29\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTransmission of Digital Information\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eNetwork security\u0026nbsp;(H04L63) ; Cryptography and security protocols (H04L9) ; Network services or applications (H04L67) ; User-to-user messaging (H04L51)\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\u003eG06F (740.47) ; G01R (3.77) ; G01M (3.73)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e41.66\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eElectrical Digital Data Processing\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003ePattern recognition (G06F18) ; Information retrieval; Database structures therefor; File system structures therefor (G06F16) ; Handling natural language data (G06F40) ; Computer-aided design [CAD] (G06F30) ; Protection from unauthorised activity (G06F21)\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\u003eA61B (646.8) ; G16H (223.57) ; A61N (13.26)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e5.581\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eHealthcare-related Diagnosis, Surgery and Identification ; Healthcare informatics\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMeasuring for diagnostic purposes (A61B5) ; Apparatus or devices for radiation diagnosis and radiation therapy equipment (A61B6) ; Diagnosis using ultrasonic, sonic or infrasonic waves (A61B8) ; ICT for medical diagnosis, simulation data mining, epidemics or pandemics (G16H50) ; Handling or processing of medical images (G16H30)\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\u003eThe definitions of the representative CPC subclasses indicate that generative networks are primarily combined with other computational models (Cluster 4), mostly neural networks and machine learning technologies, to offer utilities and address various tasks. Beyond technologies of this type, the remaining major fields concern data processing (Cluster 9) and image analysis (Clusters 1 and 7). It should be noted that the main difference between the subclasses G06T (Cluster 1) and G06V (Cluster 7) is that the former one concerns data processing and transformations. At the same time, the latter one is closer to pattern recognition and machine learning. The rest of the clusters suggest the involvement of generative networks in subjects that are not directly related to AI as well. These clusters concern healthcare (Cluster 10), vehicle and traffic control systems (Cluster 2), and business ICTs (Cluster 6). In contrast to the aforementioned areas of interest, Clusters 5 and 8 can be seen as types of methods rather than applications, meaning that generative networks are combined with data transmission systems (Cluster 8) and pictorial communication technologies (Cluster 5) in various services. Lastly, while Cluster 3 highlights the significance of speech analysis with the use of generative networks, its relatively small prevalence demonstrates the interest of organizations towards image analysis rather than audio-music data. In total, these outcomes confirm the relevance of generative AI to image and audio data as well as its adoption in diverse applications, showing that organizations have already turned their attention to technologies of this type to improve their services and internal procedures.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e4.2. Technology convergence (RO1)\u003c/h2\u003e \u003cp\u003eThe previous section showed that there are at least 10 distinct technological directions associated with generative networks, as expressed by CPC subclasses. However, more specific technological associations can be described through pairs of CPC subclasses with strong convergence. Figure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e presents the convergence network between the observations that occurred in at least 20 patent families. The subclass G06N was excluded from the analysis since it occurred in more than 20000 families. The size of each node is analogous to the number of patent families that each subclass occurred in, i.e., larger nodes correspond to more frequent observations.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe outcomes of the proposed approach suggest that there are three main nodes representing multiple convergence opportunities that can be studied for future projects and innovations. Undoubtedly, digital data processing methods (G06F) constitute the most necessary subclass to cover a variety of tasks related to generative networks. Despite the large prevalence of the CPC subclasses G06T and G06V, the CPC subclass G06F had a notably higher indegree, i.e., 42 against 15 and 10, respectively.\u003c/p\u003e \u003cp\u003eAdditionally, there are some smaller communities formed by the representative CPC subclasses of the major technology clusters (Section 4.1), more particularly G16H and A61B, H04L, G10L, and G06Q. Thus, through the above network, it is evident that while the investigated patent families can be grouped into 10 large technology clusters, more specific use cases also exist. In total, 86 connections (edges) were created, highlighting the existence of technology convergence concepts and potential opportunities covering a wide range of technological fields.\u003c/p\u003e \u003cp\u003eBeyond the primary structure of the convergence network, it is worth highlighting some of the major convergence observations that reflect both high II evaluations and co-occurrence frequencies. Starting from G06F, the most notable opportunities concern the inclusion of digital data methods in business ICTs (G06Q), systems concerning data transmission mechanisms (H04L), speech analysis (G06F), measuring electric variables (G01R), and communication networks (H04W). Other major convergence opportunities are related to applications regarding image processing-generation (G06T) and image recognition techniques (G06V); image processing-generation (G06T) and pictorial communication (H04N); image processing-generation (G06T) and healthcare-related technologies (A61B); image processing-generation (G06T) and material analysis (G01N).\u003c/p\u003e \u003cp\u003eThese insights indeed confirm multiple pathways that can inspire and guide any researcher in future projects, as proposed by the industrial and academic activity represented by patent data. As a use case, the above network can also be studied according to one or multiple target CPC subclasses to identify implementations that can elevate the potential of a complex technology or include necessary types of systems and methods represented by prevalent CPC subclasses.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e4.3. Trend Analysis (RO2)\u003c/h2\u003e \u003cp\u003eTo recognize temporal trends and the evolution of generative networks, it is worth evaluating the different states of associated technological directions over time. Below, the annual frequencies of new patented inventions related to generative networks are presented (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e), where it should be noted that only 44 new patent families were disclosed prior to 2016. The latest application date in the dataset was in January 2026, while the most recent earliest filing date among patent families was in December 2025. The analysis proves that there is indeed a significant amount of new patent families after 2022, probably because of the disclosure of new generative AI tools offering advanced capabilities and use cases.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFurthermore, Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e demonstrates the annual patent share of the 10 technology clusters, showing yearly changes and the most prolific years per technology. To provide a comparable view, the evaluations regarding yearly patent share were scaled to sum to 1 for each cluster using min-max scaling. For similar reasons, only the most significant CPC subclasses per cluster were displayed, as they cover the majority of each cluster\u0026rsquo;s information.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe evaluations indicate the emergence and decline of distinct technology clusters during the investigated time period, as both positive and negative trajectories are observed. Although neural networks and machine learning (G06N) and speech analysis techniques (G10L) peaked in 2019, they are characterized by a declining trajectory till 2023, when their annual shares stabilize. Similarly, technologies concerning image analysis and generation (G06T) were significantly developed after 2017, but also experienced similar progress after 2021. In contrast, while both digital data processing (G06F) and image pattern recognition (G06V) were less significant till 2022, they have been in demand ever since. Especially, the overall frequency of data processing technologies increased by more than 100% from 2022 to 2025.\u003c/p\u003e \u003cp\u003eMoving towards minor concepts, the evaluations suggest the recent rise of generative networks in business technologies, indicating their potential in tasks regarding administration and management (G06Q10) and commerce (G06Q30). The remaining technology clusters reached their highest patent share before 2022, highlighting the decline of related fields and the emergence of others. According to patent share, technologies related to data processing (G06F), image pattern recognition (G06V), and business tasks (G06Q) can be characterized as the only trending patent topics, since they are the only technology clusters that show a significant rise after 2022.\u003c/p\u003e \u003cp\u003eThe yearly technology share was further used to gain a deeper understanding of the growth of each technology cluster with respect to its overall prevalence instead of the yearly patent frequency (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e). In contrast to the previous evaluations, technology share suggests that the overall prevalence of most technology clusters grew significantly after 2022, but some emerged more rapidly. This is the main reason behind the consistent decrease in patent share of many technology clusters (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e). Nonetheless, technologies related to healthcare (A61B) and vehicles (B60W) exhibited the most significant decline, as their patent share peaked in 2020 and decreased thereafter. In summary, the analysis suggests that while there were distinct technology directions that caught the attention of organizations and practitioners over time, the use and orientation of generative networks did not shift from fundamental concepts, e.g., neural networks, machine learning, and image and speech analysis.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTransitioning to more specific subjects covered by individual CPC subclasses, Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e provides the technologies characterized with the highest and lowest self-growth evaluations, along with their total frequency in the entire dataset. Starting from trending technologies, the outcomes suggest that patents related to generative networks have recently included functionalities that consider energy efficiency, climate change mitigation, and pollution (Y02T, Y02D, Y02A, Y02P, and Y04S). The current analysis confirms and highlights the significance and necessity of these domains in modern implementations, which are generally regarded as mandatory for a green transition. Another arising subject is the use of generative networks in technologies concerning measuring and testing purposes (G01D, G01M, G01J, G01W, and G01H). Lastly, the remaining CPC technologies relate to more concrete subjects, indicating the recent emergence of IoT (G16Y), graphical data reading (G06K), and power generation applications with the use of generative AI.\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\u003eTrending and declining types of technologies\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\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 \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e \u003cp\u003eMost Trending Technologies\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c6\" namest=\"c4\"\u003e \u003cp\u003eMost Declining Technologies\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCPC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSelf-growth\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFrequency\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eCPC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eSelf-growth\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eFrequency\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eY02T\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e33.659\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1525\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eG03F\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.281\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e146\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eY02D\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e24.357\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e355\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eG05D\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.283\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e254\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eY02A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e16.226\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e534\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eA61N\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.288\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e143\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eY02P\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e15.677\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e517\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eH10P\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.325\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e106\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eG01D\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e11.889\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e116\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eB60R\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.343\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e47\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eG16Y\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e11.364\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e136\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eH04R\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.463\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e60\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eG01M\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e10.227\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e247\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eA61C\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.533\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e69\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eY04S\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e10.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e308\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eA63B\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.609\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e103\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eY02E\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e9.375\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e83\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eA63F\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.620\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e371\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eG06K\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e9.125\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e81\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eG10H\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.634\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e379\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eG01J\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e7.571\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e60\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eG11B\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.652\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e76\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eH02S\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e7.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e32\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eG05B\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.669\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e489\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eY10S\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4.200\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e26\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eB60L\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.744\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e136\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eG01W\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eH03M\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.756\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e72\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eG01H\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2.750\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e45\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eE21B\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.789\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e68\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\u003eOn the other end, the evaluations regarding technologies that gained less attention than others in recent years further prove the existence of declining themes as well, since the lowest observed self-growth is close to 0.28 while the highest is close to 34. In this case, the CPC subclasses could not be grouped according to similar encodings and characteristics. This means that many diverse topics constituted temporal trends that potentially did not align with the utilities of generative networks, thus not experiencing significant progress when other types of technologies were brought to the surface.\u003c/p\u003e \u003cp\u003eIn summary, the extremely high evaluations show the consistent transition of technologies into new horizons, ideas, and opportunities that should be monitored by researchers intending to develop innovative solutions. Given the findings of the analysis, these solutions should not be directed to outdated subjects but to modern methods while taking into account existing principles like climate mitigation.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e4.4. Geographical Overview (RO3)\u003c/h2\u003e \u003cp\u003eThrough a different perspective, the current section offers information about country-wise technological patterns. Jurisdictions that occurred in 200 (close to 1%) and more patent families were investigated to address only meaningful patterns and not random events. It should be mentioned that only two of the remaining concurrent jurisdictions were observed in over 50 patent families, and another three over 20. Below, Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e provides the coverage of each jurisdiction per cluster, along with the total frequency of each jurisdiction.\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\u003eJurisdiction shares per technology cluster\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"12\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c10\" colnum=\"10\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c11\" colnum=\"11\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c12\" colnum=\"12\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eJurisdiction\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFrequency\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eG06T\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eB60W\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eG10L\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eG06N\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eH04N\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eG06Q\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003eG06V\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c10\"\u003e \u003cp\u003eH04L\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c11\"\u003e \u003cp\u003eG06F\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c12\"\u003e \u003cp\u003eA61B\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eChina\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.721\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.777\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.549\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.591\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.667\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.628\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.639\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.850\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.606\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e0.711\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e \u003cp\u003e0.580\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUSA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.279\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.353\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.765\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.571\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.426\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.506\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.249\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.231\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.430\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e0.286\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e \u003cp\u003e0.524\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKorea\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.101\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.123\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.246\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.209\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.095\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.276\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.196\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.072\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.112\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e0.134\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e \u003cp\u003e0.200\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEurope\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.133\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.312\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.210\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.162\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.287\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.096\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.089\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.200\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e0.101\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e \u003cp\u003e0.254\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eJapan\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.054\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.079\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.140\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.138\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.085\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.178\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.046\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.053\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.062\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e0.052\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e \u003cp\u003e0.164\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGermany\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.024\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.039\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.201\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.037\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.038\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.033\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.005\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.027\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.014\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e0.019\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e \u003cp\u003e0.029\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCanada\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.013\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.018\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.012\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.023\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.024\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.023\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.030\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.012\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.024\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e0.015\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e \u003cp\u003e0.079\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAustralia\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.012\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.021\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.031\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.023\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.019\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.030\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.029\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.013\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.025\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e0.015\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e \u003cp\u003e0.061\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGreat Britain\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.011\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.023\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.033\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.016\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.019\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.041\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.005\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.010\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.019\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e0.009\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e \u003cp\u003e0.026\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTaiwan\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.010\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.021\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.015\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.015\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.034\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.005\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.009\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.024\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e0.010\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e \u003cp\u003e0.012\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\u003eThe outcomes show the existence of distinct countries with high involvement, marking their significant technological status and overall development compared to other countries. Undoubtedly, most innovations are protected in China, which was related to more than 70% of the patent families, while the second most prevalent country (USA) covered less than 30%. The high coverage of Chinese applications affects the significance of other countries since they dominate in all major technology clusters. For each jurisdiction, neural networks and machine learning classifications (G06N) cover a large proportion of patent classifications except for Korean patents, where technologies of this type are not as frequent.\u003c/p\u003e \u003cp\u003eMoreover, Chinese patents include fewer CPC subclasses related to this cluster on average. Overall, the relative frequencies indicate that investors operating in China are more focused on image data (G06V and G06T) and data processing methods (G06F) than other areas of interest, showing a clear direction toward providing image analysis tools. In fact, technologies associated with speech analysis (G10L), vehicles (B60W), and healthcare (A61B) were the only ones where USA applications had coverage, as Chinese applications did.\u003c/p\u003e \u003cp\u003eWhen considering the total frequency of jurisdictions, inventions related to vehicles (B60W) are notably more frequent in Germany than in any other jurisdiction, while a large proportion of the related patents are protected in the USA. Regarding speech analysis (G10L), the USA, Korea, Europe, and Japan patents constitute the most relevant to technologies of this type. Similarly, healthcare-related inventions (A61B) and technologies related to businesses (G06Q) were more developed in Canada and Australia than expected, while Korean applications included the latter types of technologies more frequently than other countries as well. Other notable evaluations concern Taiwan and Great Britain, which are closer to image processing and generation (G06T) and pictorial communication (H04N) technologies, while Taiwan is closer to network systems (H04L) than the rest of the jurisdictions as well.\u003c/p\u003e \u003cp\u003eIn summary, the primary item of interest was to demystify whether there are economic opportunities and available resources that are driven by country systems that affect the investments and involvement of third parties interested in generative networks. While the outcomes show that most inventions are protected in China, the orientation of the organizations operating in the various countries is quite different. This finding suggests that the culture, background, resources, and economic system of a country can affect domestic activities, reflecting the attraction and establishment of significant organizations, investors, and individual concepts.\u003c/p\u003e \u003cp\u003eWhen it comes to practical contributions, the current section offered information about leading applicants from a geographical perspective, where it was proved that China dominates the global market and the USA is the second most significant country. The landscape of the different countries indicated the connection between geographical characteristics and technologies surrounding generative networks. These characteristics can drive opportunities with high potential at the national level because of a suitable geographical position, primary sector, high-end technological systems, education systems, traditions, and workforce. Finally, evaluating the expertise and gaps within countries can guide future researchers in gathering information about the existing market and designing appropriate business plans.\u003c/p\u003e \u003c/div\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003eGenerative AI changed the landscape of industrial innovations and global investments by offering real-time solutions for generating useful content according to user needs and purposes. Through a patent analysis reviewing industrial developments and technological patterns, the current study offered insights into technological patterns reflecting the potential of generative networks and their overall development across time. The analysis was focused on the CPC classifications used to describe the characteristics of patents, CPC subclasses to be more precise, setting the basis towards addressing major technological directions and convergence opportunities (RO1), recent trends (RO2), and geographical patterns (RO3).\u003c/p\u003e \u003cp\u003eA total of 309 technologies were summarized into 10 clusters covering image and audio data, data processing methods, business tasks, healthcare-related implementations, vehicle systems, as well as transmission and communication technologies. These types of technologies were highlighted as the primary use cases and methods associated with generative networks according to patent data. By identifying strong associations between individual types of technologies, the paper also demonstrated a convergence network including 86 convergence opportunities that were previously observed in patent data related to generative networks.\u003c/p\u003e \u003cp\u003eFurthermore, a later trend analysis approach was followed to evaluate the progress of the various to classify trending and declining concepts since technologies evolve. This approach indicated pathways for providing up-to-date services and products and responding to the market needs as indicated by existing patented technologies. Fields related to data processing methods, business ICTs, and image pattern recognition were classified as the most trending major concepts. A complementary evaluation process also led to discovering the recent significance of power efficiency, climate change, and measuring-testing techniques, thus establishing new technological requirements for efficient and green technologies.\u003c/p\u003e \u003cp\u003eIn the last part of the paper, the main item of interest was to clarify any potential geographical patterns that would give clues about the technological status and economic activity of distinct countries. More than 70% of the innovations were covered in China, while the USA was the second most significant country involved in almost 30% of the related patent families. Despite the low frequency of other countries, the investigated patent data led to the identification of the diverging technological orientations across countries, pinpointing the existing developments of vehicle control systems in Germany, the high frequency of technologies related to image data in China, the diversity of Korean applications, and other significant findings.\u003c/p\u003e \u003cp\u003eAll in all, the current study can offer a basis for technological patterns related to generative networks from different perspectives. Essential concepts and convergence opportunities were discovered and discussed through cluster analysis and association rules, providing information regarding complex technological interconnections that demonstrate the overall potential and use cases of generative networks. The discovery of the latest trends is valuable knowledge for adapting and empowering projects with new ideas, utilities, and improvements with up-to-date technologies. Geographical patterns can be considered a reflection of the industrial and academic activities within a country, which concern available resources, technological status, and investment prospects.\u003c/p\u003e \u003cp\u003eWhen it comes to future work, the current study was dedicated to all types of technologies associated with generative AI. However, individual use cases like business ICTs can be further investigated to complement the findings provided in the current study. Furthermore, patent data includes information about the main inventors and applicants related to each patent. By analyzing this type of information, future projects can develop approaches to develop collaboration networks or address the task of competitor analysis by comparing patent portfolios. Lastly, comparing patent data with other resources, including academic literature and industrial data, is also a noteworthy idea that can reveal gaps and opportunities between sectors.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eFunding:\u003c/h2\u003e \u003cp\u003eNo funding was received for conducting this study.\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eC.K. authored the main manuscript, performed data collection and analysis, and designed the figures and table layouts.\u003c/p\u003e\u003ch2\u003eAcknowledgments:\u003c/h2\u003e \u003cp\u003eNot Applicable.\u003c/p\u003e\u003ch2\u003eAvailability of data and materials:\u003c/h2\u003e \u003cp\u003eThe data can be publicly accessed through the Patent Lens Database.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eSengar, S. S., Hasan, A. B., Kumar, S., \u0026amp; Carroll, F. (2025). Generative artificial intelligence: a systematic review and applications.\u0026nbsp;Multimedia Tools and Applications,\u0026nbsp;84(21), 23661-23700.\u003c/li\u003e\n \u003cli\u003eMittal, U., Sai, S., Chamola, V., \u0026amp; Sangwan, D. (2024). A comprehensive review on generative AI for education.\u0026nbsp;Ieee Access,\u0026nbsp;12, 142733-142759.\u003c/li\u003e\n \u003cli\u003eShokrollahi, Y., Yarmohammadtoosky, S., Nikahd, M. M., Dong, P., Li, X., \u0026amp; Gu, L. (2023). A comprehensive review of generative AI in healthcare.\u0026nbsp;arXiv preprint arXiv:2310.00795.\u003c/li\u003e\n \u003cli\u003eShafiee, S. (2025). Generative AI in manufacturing: a literature review of recent applications and future prospects.\u0026nbsp;Procedia CIRP,\u0026nbsp;132, 1-6.\u003c/li\u003e\n \u003cli\u003eKar, A. K., Varsha, P. S., \u0026amp; Rajan, S. (2023). Unravelling the impact of generative artificial intelligence (GAI) in industrial applications: A review of scientific and grey literature.\u0026nbsp;Global Journal of Flexible Systems Management,\u0026nbsp;24(4), 659-689.\u003c/li\u003e\n \u003cli\u003eFujii, H., \u0026amp; Managi, S. (2018). Trends and priority shifts in artificial intelligence technology invention: A global patent analysis.\u0026nbsp;Economic Analysis and Policy,\u0026nbsp;58, 60-69.\u003c/li\u003e\n \u003cli\u003eAbadi, H. H. N., \u0026amp; Pecht, M. (2020). Artificial intelligence trends based on the patents granted by the United States patent and trademark office.\u0026nbsp;IEEE Access,\u0026nbsp;8, 81633-81643.\u003c/li\u003e\n \u003cli\u003eLiu, N., Shapira, P., Yue, X., \u0026amp; Guan, J. (2021). Mapping technological innovation dynamics in artificial intelligence domains: Evidence from a global patent analysis.\u0026nbsp;Plos one,\u0026nbsp;16(12), e0262050.\u003c/li\u003e\n \u003cli\u003eChang, S. H. (2021). Technical trends of artificial intelligence in standard-essential patents.\u0026nbsp;Data Technologies and Applications,\u0026nbsp;55(1), 97-117.\u003c/li\u003e\n \u003cli\u003eGiczy, A. V., Pairolero, N. A., \u0026amp; Toole, A. A. (2022). Identifying artificial intelligence (AI) invention: A novel AI patent dataset.\u0026nbsp;The Journal of Technology Transfer,\u0026nbsp;47(2), 476-505.\u003c/li\u003e\n \u003cli\u003eAilia, M. J., Thakur, N., Abdul-Ghafar, J., Jung, C. K., Yim, K., \u0026amp; Chong, Y. (2022). Current trend of artificial intelligence patents in digital pathology: a systematic evaluation of the patent landscape.\u0026nbsp;Cancers,\u0026nbsp;14(10), 2400.\u003c/li\u003e\n \u003cli\u003eLee, M. (2020). An analysis of the effects of artificial intelligence on electric vehicle technology innovation using patent data.\u0026nbsp;World Patent Information,\u0026nbsp;63, 102002.\u003c/li\u003e\n \u003cli\u003eBenjamens, S., Dhunnoo, P., G\u0026ouml;r\u0026ouml;g, M., \u0026amp; Mesko, B. (2023). Forecasting artificial intelligence trends in health care: systematic international patent analysis.\u0026nbsp;Jmir Ai,\u0026nbsp;2(1), e47283.\u003c/li\u003e\n \u003cli\u003eAbbas, A., Zhang, L., \u0026amp; Khan, S. U. (2014). A literature review on the state-of-the-art in patent analysis.\u0026nbsp;World Patent Information,\u0026nbsp;37, 3-13.\u003c/li\u003e\n \u003cli\u003ePejić Bach, M., Pivar, J., \u0026amp; Fabac, R. (2024). Emerging trends in generative artificial intelligence: Insights from patent analysis using Lens. org toolkit.\u0026nbsp;ITIS 2024 BOOK OF PROCEEDINGS\u0026apos;\u0026apos;Harnessing the power of big data for green and digital transition\u0026apos;\u0026apos;, 151-160.\u003c/li\u003e\n \u003cli\u003eCharmanas, K., Mittas, N., \u0026amp; Angelis, L. (2023a). Topic and influence analysis on technological patents related to security vulnerabilities.\u0026nbsp;Computers \u0026amp; Security,\u0026nbsp;128, 103128.\u003c/li\u003e\n \u003cli\u003eCharmanas, K., Georgiou, K., Mittas, N., \u0026amp; Angelis, L. (2023b). Classifying the main technology clusters and assignees of home automation networks using patent classifications.\u0026nbsp;Computers,\u0026nbsp;12(10), 211.\u003c/li\u003e\n \u003cli\u003eYang, G. C., Li, G., Li, C. Y., Zhao, Y. H., Zhang, J., Liu, T., ... \u0026amp; Huang, M. H. (2015). Using the comprehensive patent citation network (CPC) to evaluate patent value.\u0026nbsp;Scientometrics,\u0026nbsp;105(3), 1319-1346.\u003c/li\u003e\n \u003cli\u003eChoi, D., \u0026amp; Song, B. (2018). Exploring technological trends in logistics: Topic modeling-based patent analysis.\u0026nbsp;Sustainability,\u0026nbsp;10(8), 2810.\u003c/li\u003e\n \u003cli\u003eChoi, H., Oh, S., Choi, S., \u0026amp; Yoon, J. (2018). Innovation topic analysis of technology: The case of augmented reality patents.\u0026nbsp;IEEE Access,\u0026nbsp;6, 16119-16137.\u003c/li\u003e\n \u003cli\u003eHu, X., Rousseau, R., \u0026amp; Chen, J. (2012). A new approach for measuring the value of patents based on structural indicators for ego patent citation networks.\u0026nbsp;Journal of the American Society for Information Science and Technology,\u0026nbsp;63(9), 1834-1842.\u003c/li\u003e\n \u003cli\u003eLee, M., \u0026amp; Lee, S. (2017). Identifying new business opportunities from competitor intelligence: An integrated use of patent and trademark databases.\u0026nbsp;Technological Forecasting and Social Change,\u0026nbsp;119, 170-183.\u003c/li\u003e\n \u003cli\u003eHaghighian Roudsari, A., Afshar, J., Lee, W., \u0026amp; Lee, S. (2022). PatentNet: multi-label classification of patent documents using deep learning based language understanding.\u0026nbsp;Scientometrics,\u0026nbsp;127(1), 207-231.\u003c/li\u003e\n \u003cli\u003eTrappey, A. J., Trappey, C. V., Govindarajan, U. H., \u0026amp; Sun, J. J. (2019). Patent value analysis using deep learning models\u0026mdash;The case of IoT technology mining for the manufacturing industry.\u0026nbsp;IEEE Transactions on Engineering Management,\u0026nbsp;68(5), 1334-1346.\u003c/li\u003e\n \u003cli\u003eKarataş, A. R., Kazak, H., Akcan, A. T., Akkaş, E., \u0026amp; Arık, M. (2024). A bibliometric mapping analysis of the literature on patent analysis.\u0026nbsp;World Patent Information,\u0026nbsp;77, 102266.\u003c/li\u003e\n \u003cli\u003eLin, T. Y., \u0026amp; Chou, L. C. (2025). A systematic review of artificial intelligence applications and methodological advances in patent analysis.\u0026nbsp;World Patent Information,\u0026nbsp;82, 102383.\u003c/li\u003e\n \u003cli\u003eShomee, H. H., Wang, Z., Ravi, S. N., \u0026amp; Medya, S. (2025, July). A survey on patent analysis: From nlp to multimodal ai. In\u0026nbsp;Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)\u0026nbsp;(pp. 8545-8561).\u003c/li\u003e\n \u003cli\u003eVerendel, V. (2023). Tracking artificial intelligence in climate inventions with patent data.\u0026nbsp;Nature Climate Change,\u0026nbsp;13(1), 40-47.\u003c/li\u003e\n \u003cli\u003eSantarelli, E., Staccioli, J., \u0026amp; Vivarelli, M. (2023). Automation and related technologies: a mapping of the new knowledge base.\u0026nbsp;The Journal of Technology Transfer,\u0026nbsp;48(2), 779-813.\u003c/li\u003e\n \u003cli\u003eHu, T., Guo, J., Zhang, T., Liu, J., Sun, X., \u0026amp; Chang, Z. (2022, June). Analysis of the development trend of artificial intelligence technology application in the field of sports-based on patent measurement. In\u0026nbsp;International Conference on Human-Computer Interaction\u0026nbsp;(pp. 337-354). Cham: Springer International Publishing.\u003c/li\u003e\n \u003cli\u003ePark, H. S. (2017). Technology convergence, open innovation, and dynamic economy.\u0026nbsp;Journal of Open Innovation: Technology, Market, and Complexity,\u0026nbsp;3(4), 1-13.\u003c/li\u003e\n \u003cli\u003eCharmanas, K., Filippou, K., Mittas, N., \u0026amp; Angelis, L. (2025a). Exploring the role of information security news descriptions on retweet proneness and user interactions.\u0026nbsp;Journal of Computational Social Science,\u0026nbsp;8(3), 60.\u003c/li\u003e\n \u003cli\u003eLee, D. D., \u0026amp; Seung, H. S. (1999). Learning the parts of objects by non-negative matrix factorization.\u0026nbsp;nature,\u0026nbsp;401(6755), 788-791.\u003c/li\u003e\n \u003cli\u003eLee, D., \u0026amp; Seung, H. S. (2000). Algorithms for non-negative matrix factorization.\u0026nbsp;Advances in neural information processing systems,\u0026nbsp;13.\u003c/li\u003e\n \u003cli\u003eKim, H., \u0026amp; Park, H. (2007). Sparse non-negative matrix factorizations via alternating non-negativity-constrained least squares for microarray data analysis.\u0026nbsp;Bioinformatics,\u0026nbsp;23(12), 1495-1502.\u003c/li\u003e\n \u003cli\u003eŽitnik, M., \u0026amp; Zupan, B. (2012). Nimfa: A python library for nonnegative matrix factorization.\u0026nbsp;The Journal of Machine Learning Research,\u0026nbsp;13(1), 849-853.\u003c/li\u003e\n \u003cli\u003eDieng, A. B., Ruiz, F. J., \u0026amp; Blei, D. M. (2020). Topic modeling in embedding spaces.\u0026nbsp;Transactions of the Association for Computational Linguistics,\u0026nbsp;8, 439-453.\u003c/li\u003e\n \u003cli\u003eZhang, S., \u0026amp; Wu, X. (2011). Fundamentals of association rules in data mining and knowledge discovery.\u0026nbsp;Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery,\u0026nbsp;1(2), 97-116.\u003c/li\u003e\n \u003cli\u003eHe, Q. (1999). Knowledge discovery through co-word analysis.\u003c/li\u003e\n \u003cli\u003eCharmanas, K., Georgiou, K., Papageorgiadis, K., Mittas, N., \u0026amp; Angelis, L. (2025b). A topic-oriented trend analysis framework for Stack Exchange questions: Case study on ChatGPT related queries on Stack Overflow. Information and Software Technology, 107969.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Footnotes","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.uspto.gov/web/patents/classification/cpc/html/cpc.html\u003c/span\u003e\u003cspan address=\"https://www.uspto.gov/web/patents/classification/cpc/html/cpc.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.lens.org/\u003c/span\u003e\u003cspan address=\"https://www.lens.org/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.epo.org/en\u003c/span\u003e\u003cspan address=\"https://www.epo.org/en\" targettype=\"URL\" 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":true,"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":"Generative AI, Patent Analysis, Statistical Analysis, Clustering, Convergence Networks","lastPublishedDoi":"10.21203/rs.3.rs-9446615/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9446615/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eTechnological novelties mark the rise of new ideas, impacting both scientific tools and industrial products, and can be monitored through public resources. Patents constitute a resource for tracking essential and breakthrough innovations to evaluate their overall development in the market. Generative Artificial Intelligence (AI) is one of these breakthroughs, offering capabilities that facilitate knowledge acquisition, writing, and image processing among many common and professional tasks. This study examines technologies related to generative networks using international patent documents from the Patent Lens Database. The Cooperative Patent Classification (CPC) system was the basis for identifying associated types of technologies and their interconnections through cluster analysis and convergence networks, while other metadata were also evaluated to address major trends and geographical patterns. CPC classifications show that generative networks are primarily combined with image processing techniques, data analysis pipelines, and computational models, while also finding implementations in other indirectly related fields concerning electric vehicles, healthcare, and business tasks. Technology convergence indicators showed multiple potential pathways that highlight the interconnections between different technologies and potential use cases. Patent application dates suggest the emergence of methods and applications concerning power efficiency, climate mitigation, measuring and testing, IoT, and graphical data reading, while the major concepts of data processing, image pattern recognition, and business implementations were classified as arising too. Evaluations regarding patent jurisdictions indicate the existence of experts and prevalent investors operating in distinct developed countries, whereas country-wise preferences are confirmed. Altogether, this study provides insights into major technological directions related to generative networks, demonstrates convergence opportunities, highlights recent trends, and addresses geographical patterns.\u003c/p\u003e","manuscriptTitle":"Patent Landscape of Generative Networks: A Data-Driven Examination of Essential Concepts Convergence Opportunities and Technological Trends Through Patent Analysis","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-24 07:18:48","doi":"10.21203/rs.3.rs-9446615/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":"ed7c88ba-3741-4c8e-8bed-12e5f72aa944","owner":[],"postedDate":"April 24th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-04-25T21:38:33+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-24 07:18:48","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9446615","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9446615","identity":"rs-9446615","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}