The Evolution of Consumer Emotions in Response to Rhetorical Strategies in Corporate Social Responsibility Advertising: Evidence from Chinese Social Networks Before and After the COVID-19 Pandemic

The Evolution of Consumer Emotions in Response to Rhetorical Strategies in Corporate Social Responsibility Advertising: Evidence from Chinese Social Networks Before and After the COVID-19 Pandemic | 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 Article The Evolution of Consumer Emotions in Response to Rhetorical Strategies in Corporate Social Responsibility Advertising: Evidence from Chinese Social Networks Before and After the COVID-19 Pandemic jianguo wang, xixiang sun, wei zhang, yige jia, you li This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5697919/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 In the post-pandemic era, the relationship between Corporate Social Responsibility (CSR) advertising and consumer emotions has become a significant research focus within the CSR field. This study employs the BERTopic model and SnowNLP sentiment analysis, utilizing the Python 3.10 system to mine data from Weibo, to deeply explore how rhetorical themes in text affect the relationship between CSR advertising and consumer emotions. The findings reveal that: (1) CSR advertising has shifted from declarative expressions to pun-based rhetorical themes, with puns becoming the primary communication strategy; (2) Influenced by the COVID-19 pandemic, the context created by CSR advertising has gradually shifted from urging immediate consumption behaviors in the "present" to advocating that current actions can have a positive impact on the "future"; (3) Consumers show a strong enthusiasm and interest in CSR advertisements that use puns for marketing communication, far exceeding that for other rhetorical strategies. Business and commerce/Business and management Humanities/Cultural and media studies BERTopic SnowNLP CSR Advertising Rhetorical Strategies Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 0 Introduction Driven by technological advancements, commercial companies are exploring innovative methods to create value by connecting with consumers and enhancing engagement and interaction (Wang,2021). As information technology evolves and social media rises, interactive marketing has become crucial for companies involved in corporate social responsibility (CSR) efforts. Higher perceived interactivity in CSR communications can enhance information credibility and strengthen consumer corporate identification, improving reputation and increasing willingness to share positive word-of-mouth(Eberle et al., 2013 ). Interactive technologies, like websites and social media platforms, make CSR communication more engaging and foster dialogue between companies and consumers (Tomaselli & Melia, 2014 ). Additionally, CSR significantly impacts corporate performance when communicated through advertising (Rahman et al., 2017 ). Advertising intensity positively moderates the relationship between CSR activities and market share(Rahman et al., 2017 ). Both advertising and CSR are essential for companies to build reputation, enhance brand authenticity, and boost credibility (Lloyd-Smith & An, 2019 ). Since its inception, research on corporate social responsibility has been closely linked with brand building, corporate advertising, and consumer emotions (Hartmann et al., 2023 )。 The global COVID-19 pandemic has significantly impacted the brand image of companies that have long positioned themselves as green, environmentally friendly, and sustainable. Additionally, the pandemic has shaken consumer confidence in CSR initiatives (Panagiotopoulos, 2021 ). This global health crisis is testing companies' commitment to CSR, with evidence suggesting that many are reevaluating and adjusting their CSR strategies to better adapt to the crisis and meet changing public expectations(Carroll, 2021 ). In response to pandemic challenges, many companies are incorporating more altruistic messaging in CSR advertising to rebuild trust and strengthen the consumer relationship with CSR initiatives(Xie & Wang, 2022). As we move into the post-COVID-19 era, it is increasingly important to explore how companies can use CSR advertising effectively to influence consumer emotions and restore confidence in CSR practices. This study aims to examine the relationship between CSR advertising and consumer emotions by investigating strategic approaches to the processing of textual information in CSR advertisements. Our study aims to investigate the emotional connections between rhetorical strategies used in CSR advertisements and consumer reactions by applying cluster analysis to CSR ads gathered from Chinese social networking platforms. This research seeks to identify specific rhetorical techniques that, when used in creating structured and curated textual content, effectively elicit positive emotional responses from consumers. By understanding these strategies, we aim to reveal how CSR advertising can foster closer, more intimate relationships between consumers and CSR initiatives. The structure of this study is as follows: First, a literature review is conducted, analyzing previous research on corporate social responsibility, advertising communication, and textual rhetorical strategies. Next, the methodology section outlines the data collection and cleaning processes. The study employs clustering analysis using the BERTopic model(Grootendorst, 2022 ), which is based on the Top2Vec framework(Angelov, 2020 ). Subsequently, SnowNLP is applied to perform sentiment analysis on the rhetorical strategies identified within the clusters(Cai, 2021 ). The paper concludes with a discussion of the findings and provides recommendations for future research. 1 Literature Review 1.1 Current State of CSR Advertising Research CSR advertising is a vital area of study, examining how companies use advertising to communicate CSR activities to consumers and build relationships. Research has explored the dual objectives of CSR advertising: enhancing corporate reputation and positively influencing consumer attitudes. A significant focus of CSR advertising research is its impact on corporate reputation. Studies show that effective CSR advertising can lead to favorable consumer perceptions of a company’s image, enhancing brand equity (Chang, 2012 ; Chernev & Blair, 2015 ). Consumers generally respond positively to CSR ads that align with their values, highlighting the importance of strategic message alignment in CSR campaigns (Diehl et al., 2016 ). Another key research area is stakeholder reactions, particularly consumer and employee responses to CSR advertising. Literature suggests that positive consumer responses often depend on the perceived authenticity and alignment between a company's CSR efforts and its brand identity (C.-W. Choi, 2022 ). Consumers are more likely to trust and support companies whose CSR activities appear genuinely committed to social good, rather than being self-serving (Bergkvist & Zhou, 2019 ). The rise of digital platforms has dramatically reshaped CSR advertising strategies. Research highlights the growing influence of social media and electronic word-of-mouth (eWOM) in CSR communication. Social media is a powerful tool for disseminating CSR messages, facilitating consumer engagement, and co-creating content, thereby amplifying the reach and impact of CSR initiatives (Schouten et al., 2020 ). Studies also emphasize the role of influencers in CSR advertising, noting that the trustworthiness and perceived authenticity of influencers shape consumer attitudes toward CSR messages(C. S. Choi et al., 2019 ). Authenticity remains a crucial theme in CSR advertising. According to the Persuasion Knowledge Model, consumers' ability to recognize and evaluate the authenticity of CSR efforts significantly influences their reactions (Friestad & Wright, 1994 ). When consumers perceive CSR advertisements as insincere or purely profit-driven, skepticism may arise, potentially damaging the brand's reputation (Jing Wen et al., 2020 ).Thus, maintaining authenticity in CSR messaging is essential to prevent consumer backlash. 1.2 Current Research on Textual Rhetorical Strategies Language is a crucial tool in marketing communication. As early as 1923, American communication pioneer Claude C. Hopkins noted in Scientific Advertising that copywriting rhetoric, sentence structure arrangement, and headline wording strongly appeal to consumers. In today's interactive marketing environment, understanding consumers' linguistic psychology helps green marketers communicate more effectively and reach consumers' minds swiftly., For example, green brands like McDonald's and Starbucks have captured consumer attention, sparked discussion, and encouraged active sharing through eco-friendly content on Chinese social media platforms. The use of language in marketing communication has been a long-standing area of interest. Numerous studies have applied psycholinguistic paradigms and methods to marketing contexts(Lowrey, 2002 ).These studies show that language elements in marketing—from brand naming and slogans to copywriting rhetoric, advertising voiceovers, and product descriptions—significantly impact consumer memory, attitudes, and behavior(Pogacar et al., 2018 ) .In marketing, textual language functions to transmit information, convey values, explain, and persuade(Luangrath et al., 2017 ), Therefore, existing research primarily focuses on how different rhetorical uses in text affect consumer emotional responses. Recent research shows a close relationship between textual rhetorical strategies and CSR..CEOs use various rhetorical methods in CSR communication, such as ethos, pathos, and logos, where pathos and logos positively impact corporate social performance(Liu et al., 2019 ). Three primary categories of rhetoric are identified: value rhetoric for moral legitimacy, normative rhetoric for cognitive legitimacy, and instrumental rhetoric for pragmatic legitimacy (Marais, 2012 ). An analysis of CSR reports shows that ethos is the most commonly used moral perspective, followed by benefit and agency(Ditlev-Simonsen & Wenstøp, n.d.). Companies employ these rhetorical strategies in CSR reports to manage stakeholder expectations and legitimize their actions(Feldner, 2014 ). The most prevalent pattern links agency and benefit perspectives, suggesting that the pursuit of profit is justified by agency. These findings highlight the importance of rhetorical strategies in CSR communication and their impact on stakeholder perceptions. Current academic research on the relationship between CSR and rhetorical strategies predominantly centers on benefit-related perspectives. However, limited research has explored the connection between the specific wording used in texts and CSR. Therefore, this study seeks to uncover the relationship between textual rhetoric in CSR advertisements and consumer emotions. 1.3 State of the Art in Thematic Modeling Research Latent Dirichlet Allocation (LDA) is a generative statistical model commonly used for topic modeling to discover abstract topics in document collections (Blei, Ng, & Jordan, 2003). LDA assumes each document is a mixture of a few topics, with each word attributable to one of these topics. Since its introduction, LDA has become a standard tool for text analysis, valued for its robustness in handling large datasets and its capacity to uncover hidden thematic structures. LDA has been widely applied in CSR research to analyze and understand various dimensions of CSR activities and their communication. Several key studies have demonstrated LDA's effectiveness in analyzing CSR-related content. For instance, Lee and Carroll (Lee & Carroll, 2011 ) used LDA to examine CSR reports from Fortune Global 500 companies, identifying key themes such as environmental management, employee welfare, and ethical governance. Another study by Gao, Yu, and Lee(Gao, 2011 ) applied LDA to analyze CSR disclosures in China, highlighting how CSR themes evolved over time. However, LDA, which relies on the bag-of-words approach, often fails to capture semantic relationships between words.. Because this method ignores the context of word appearance, it cannot accurately represent document content. To address these limitations, Maarten Grootendorst introduced the BERTopic model in 2020, which better captures the intricate relationships within text. Research in topic modeling has seen continuous advancements tailored to different textual characteristics. Traditional LDA models often face data sparsity issues, particularly in short texts, where sparse data limits reliable extraction of word co-occurrence patterns. Therefore, this study utilizes the BERTopic model to overcome limitations associated with traditional LDA topic modeling. The BERTopic model offers enhanced semantic interpretability, enabling it to better understand semantic connections and contextual information within text data, thereby more precisely capturing the underlying meaning of topics. Sentiment analysis is a widely used technique in social media data analysis, with methods broadly categorized into dictionary-based and machine learning-based approaches. Dictionary-based sentiment analysis uses pre-constructed sentiment lexicons to analyze emotions in text, such as the Hownet Sentiment Lexicon developed by China National Knowledge Infrastructure (CNKI). In contrast, machine learning-based sentiment analysis involves training models to recognize and classify emotions from data, often offering higher accuracy. Examples include using Recurrent Neural Networks (RNNs) and Convolutional Neural Networks (CNNs) for sentiment analysis. Some studies have used pre-packaged third-party sentiment analysis libraries to conduct sentiment analysis. Yu et al. utilized SnowNLP to detect emotions on Chinese social media platforms (Weibo) during the COVID-19 pandemic(Yu et al., 2021 ) .Chinese scholars, Zhang et al., used SnowNLP to conduct sentiment analysis and monitor concerns about COVID-19 vaccines on Weibo(Zhang et al., 2021 ).Researchers used SnowNLP and descriptive statistics to analyze social networks, demonstrating interactions among different participants, and used social media analysis to assess the impact of Weibo on public emotions(Xu et al., 2023 )。Chinese scholar Hu(Hu, 2022 ) utilized SnowNLP to analyze the emotional trends of online public opinion during major epidemics. Sentiment analysis, as an important tool in text analysis, requires careful selection based on different text analysis scenarios. Different sentiment analysis tools have their own strengths and applicable scenarios. In the specific context of social media sentiment analysis, the widespread use of SnowNLP has demonstrated its superior performance. Therefore, this study selects SnowNLP as the tool for sentiment analysis. 2 Research Objective This study aims to address the following primary research objectives (RO): By analyzing semantic dissemination texts on social media platforms, we seek to explore the relationship between textual rhetorical strategies used in CSR advertisements and consumer emotions, thereby establishing large-scale textual comparisons in this expanding field of research. Therefore, this study specifically proposes the following research objectives: RO.1: By preprocessing (cleaning data) and applying data mining techniques to a large volume of CSR advertisement texts from social media platforms, this study aims to compare the types of rhetorical themes used in CSR advertising content before and after the COVID-19 pandemic. RO.2. By clustering CSR advertisement content from before and after the COVID-19 pandemic and extracting thematic keywords to identify their moral context, this study explores the relationship between CSR advertising themes and their moral context. RO.3. To explore the impact of rhetorical strategies used in CSR advertisements on consumers, this study seeks to clarify the relationship between the rhetorical themes of CSR advertising and consumer emotions. Our research reveals the following: (1) it overcomes the limitations of probabilistic generative topic models; (2) it expands upon previous research on text processing and semantic themes; and (3) it applies exploratory methods using large-scale data sources. Therefore, compared to previous studies that relied on limited sample data, this research offers greater reliability and accuracy. 3 Research Design As illustrated in Figure 1, this study's research framework for theme-emotion integration in CSR advertising rhetorical strategies starts with using Python 3.10 to collect target textual data from the Weibo platform, followed by data preprocessing and classification. Next, the BERTopic model is used to identify relevant thematic content in CSR advertising rhetorical strategies across different periods, before and after the COVID-19 pandemic. Additionally, SnowNLP analyzes the emotional evolution of consumer comments, revealing the emotional impact of CSR advertising rhetorical strategies on consumers. 3.1 Data Collection This study examines the rhetorical content of CSR advertisements. Recognizing that environmental activities can influence consumers' identity associations(H.-J.Wang, 2019) , we selected topics like #World Water Day (March 22), #World Earth Day (April 22), and #World Environment Day (June 5) to emphasize CSR identity. We then collected related texts from the Weibo platform. Data collected included user IDs, posting times, post content, number of likes, and comment texts. Our analysis was restricted to Simplified Chinese text data to ensure consistency. The study utilized Python 3.10 data packages. To ensure scientific rigor and data validity, we adopted Sánchez-Franco and Rey-Moreno's(Sánchez‐Franco & Rey‐Moreno, 2022) framework, dividing data collection into two periods: pre-COVID-19 (March 22, 2017 - June 5, 2019) and post-COVID-19 (March 22, 2021 - June 5, 2023). As a result, the dataset comprised 834 microblog posts and 109,474 comments. 3.2 Data Cleaning During the data cleaning phase, many texts contained terms related to specific interactive marketing activities, such as "World Earth Day," "World Water Day," and "World Environment Day.". To enhance segmentation accuracy, we built a custom dictionary. While the BERTopic model imposes no restrictions on sample data, based on Sánchez‐Franco and Rey‐Moreno（2022) , we aimed to better extract relevant text features, reduce dataset dimensions, and ensure data consistency.(1)We removed duplicate data;(2)Excluded non-brand-generated content;(3)Deleted common stop words to filter out frequent terms and customize proper nouns;(4)Employed jieba for text segmentation;(5)Eliminated noise text;(6)Tokenized symbols and emojis in comments. In the BERTopic model, each sentence is considered an independent text unit. Therefore, we applied sentence-level processing to more effectively capture subtle differences. The final dataset consisted of 526 pieces of green brand-generated content and 69,312 comments. 3.3 BERTopic Thematic Model Analysis This study uses the BERTopic model for theme identification via text clustering and then conducts sentiment analysis on Weibo posts within each identified theme. This method supports the analysis of rhetorical strategies in CSR advertising content. The BERTopic model is a natural language processing technique based on BERT (Bidirectional Encoder Representations from Transformers) that automatically constructs topic models from large volumes of text data. In BERTopic, textual content is generally viewed as consisting of related topics and associated words. While a text may encompass multiple topics, each word is assigned to a specific topic. Compared to traditional topic mining methods, BERTopic provides greater insight into latent semantics, particularly excelling in processing short texts, outperforming methods such as Latent Dirichlet Allocation (LDA) and Probabilistic Latent Semantic Analysis (PLSA). This study begins by using Sentence-BERT for pre-training, generating text embedding vectors for each sentence with a pre-trained BERT model. This process produces a similarity score between -1 and 1, where 1 indicates complete similarity and -1 indicates complete dissimilarity. By vectorizing the text in this way, the vectors of two sentences can be concatenated to measure their similarity. Sentence-BERT captures the underlying semantic relationships between sentences using this method(Reimers & Gurevych, 2019). UMAP (Uniform Manifold Approximation and Projection) is a highly efficient nonlinear dimensionality reduction technique, primarily designed for handling large-scale datasets. The algorithm constructs a local neighborhood graph of the original high-dimensional data, seeking an optimal topological structure in a lower-dimensional space to preserve both global and local structures effectively. The algorithm constructs a local neighborhood graph of the original high-dimensional data, seeking an optimal topological structure in a lower-dimensional space to preserve both global and local structures effectively. In practice, UMAP is widely applied in data visualization, clustering analysis, feature extraction, and anomaly detection. Its flexibility and scalability make it a versatile tool for dimensionality reduction in modern data science and machine learning. (McInnes et al., 2018). HDBSCAN (Hierarchical Density-Based Spatial Clustering of Applications with Noise) is an advanced clustering algorithm that identifies data clusters with varying densities and shapes. It performs hierarchical clustering by constructing a minimum spanning tree (MST), which allows it to automatically determine the number of clusters and manage noise effectively. Its robustness enables it to accurately detect primary clustering patterns, even in the presence of outliers and noise.(McInnes & Healy, 2017). Following the application of UMAP and HDBSCAN, we employed the Class-based TF-IDF method to extract keywords. Class-based TF-IDF is a variant of the traditional term frequency-inverse document frequency (TF-IDF) method, specifically designed to improve text classification effectiveness by incorporating class information into the calculation. A higher Class-based TF-IDF value indicates a more representative term, making it suitable as a keyword for clustering themes. Based on several experiments and comparisons, our study determined that the optimal number of research topics is seven. (Egger & Yu, 2022). Finally, Top2Vec was employed to merge the most similar topic vectors, reducing the number of topics.Top2Vec combines semantically related topics, effectively reducing the number of topics while preserving semantic relevance. This approach ensures a suitable range of topics by merging them without compromising semantic integrity. Top2Vec accomplishes this by calculating the similarity between topic vectors, identifying the most similar pairs, and merging them into new topic vectors. This process provides optimal topic categorization for the overall BERTopic analysis (Egger & Yu, 2022). This study focuses on the types of rhetorical strategies used in CSR advertising content to explore their relationship with consumer emotions. However, rhetoric is rooted in linguistic systems structured by human activities(Taboada & Mann,2006). Therefore, this study employs machine learning techniques to cluster datasets by different data collection phases, identifying the most cohesive thematic sentences based on topic dispersion. Subsequently, three linguistics experts were recruited to annotate the samples based on the definitions of rhetorical figures provided in the Modern Chinese Dictionary (Li et al., 2024). Due to the overlap among the 16 rhetorical figures currently classified in Chinese rhetoric, metaphor, simile, and metonymy were grouped into a single category called metaphorical rhetoric to facilitate clustering analysis. The experts independently coded the samples, and in cases of disagreement, they discussed to reach a consensus on each text's classification (Huang et al., 2022). After annotation, four main clustering themes emerged: three thematic models using rhetorical strategies—pun, rhetorical question, and metaphor—and one non-rhetorical model, the statement. 3.4 Sentiment Analysis SnowNLP is a tool specifically designed for processing Chinese text, implemented on the Python platform, making it particularly suitable for handling Simplified Chinese. It supports various natural language processing functions, including sentiment analysis, word segmentation, part-of-speech tagging, keyword extraction, text classification, and converting text to pinyin. For sentiment analysis, the Chinese text is first passed to the SnowNLP module. SnowNLP examines the words, word relationships, and sentence structures to check for matches in its sentiment dictionary and identify any negative expressions. Once a match is found, the sentiment polarity of the words (positive, neutral, or negative) is determined, and the counts of each type of sentiment word are calculated. Based on the sentiment word counts, their polarity, and the contextual semantics, SnowNLP computes a sentiment score to indicate the emotional intensity of the text. According to the scoring criteria, a score below 0.3 indicates negative sentiment, between 0.3 and 0.6 indicates neutral sentiment, and above 0.6 indicates positive sentiment. SnowNLP then outputs the sentiment score of the text based on this analysis (Chen et al., 2018). 3.5 Theme-Sentiment Evolution Analysis This study performs an evolutionary analysis of the rhetorical strategies in CSR advertising text content across two dimensions: theme clustering and sentiment relationships. A theme-sentiment evolution analysis is then conducted based on this framework. By comparing the distribution of themes over different data collection periods, we can map the temporal evolution trajectories of these themes and examine how they relate to sentiment evolution. A Sankey diagram is employed to visualize the results. 4 Results Analysis 4.1 Theme Analysis 4.1.1 Theme analysis before the COVID-19 pandemic. Before the COVID-19 pandemic outbreak, 262 Weibo posts and 34,525 comments were collected using Python 3.10. The BERTopic model was employed to extract themes during this period. To enhance interpretability, theme description accuracy, and refine topic modeling results, language experts were invited to summarize and categorize the themes. After summarization, 21 clustered thematic sentences emerged in the pre-pandemic stage, categorized into four types: rhetorical questions, metaphorical rhetoric, puns, and statements (Table1). Statements constituted the majority, with 11 themes accounting for 52.5% of the total, indicating that CSR advertisements predominantly used declarative expressions for consumer communication before the pandemic. Puns and rhetorical questions each had four themes, representing 19% of the total, suggesting that a minority of CSR advertisements used these rhetorical strategies to engage consumers. Finally, metaphorical rhetoric had only two themes, making up 9.5%, indicating that businesses were less inclined to use metaphors for communicating with consumers in CSR advertisements. Analysis of thematic keywords using the BERTopic model (Figure 2) reveals that a considerable portion of the content revolves around temporal contexts like "the present," "now," and "today." This indicates that, prior to the COVID-19 pandemic, CSR advertisements mainly focused on the present aspects of people's lives, paying little attention to future thoughts and plans. Table 1: Themes Before COVID-19 Before COVID-19 Theme Types Examples of Thematic Sentences Keywords Declarative Keystone species typically refer to those that play a crucial role in maintaining biodiversity, as well as the structure, function, and stability of ecosystems or biological communities. Diversity/Ecosystem /Biological/Function Question What can we do to love the Earth? Today, share for the planet: choose low-carbon transportation, save water, reduce plastic use, clean your plate, and sort waste. Let's take action together! Today/Earth/Action Pun Protect the environment, let's take 'art'-ion together! Protect/Action/Environment Declarative Don’t let the beauty of Earth exist only in images; don’t let nature and vitality remain only in memories. Humanity should not only explore the future but also join hands to protect the present. Now/Protect/Nature Declarative Let us focus on the importance of climate change and biodiversity conservation, and raise our sense of responsibility. Let's be friends of nature together. Climate Change/Biological/Conservation Declarative Let us respect nature, protect nature, and work together to build a community of life between humans and nature to address the challenges of climate change. Climate Change/Respect/Nature Declarative Protecting nature and caring for the environment are not just slogans for today but actions for the future. Today/Protect/Eenvironment Metaphor Let the Earth thrive with an extra layer, like a wooden coat, for healthy growth Earth/Health Question In your daily life, what small things do you consistently do for environmental protection? Daily Life/Protect Question Do you wish every drop of water could be clean? Then start by saving water. Clean/Save Declarative Committed to reducing carbon emissions and promoting green development, our innovative technologies not only change lives but also protect the Earth. Let's work together for a sustainable future! Committed to/Innovation Declarative Remember the source when you drink water, and respect nature! Baisui Mountain is always committed to protecting our water sources. Protect/Committed to Declarative Let us join hands now to protect nature and work together to slow down the trend of global warming. Now/Protect/Nature Question Did you know? Approximately 7 million people die each year from air pollution, at least 60,000 species go extinct annually, and 6 million hectares of land turn into deserts every year... Let’s take action together to protect our clear waters and green mountains! Pollution/Action/Now Declarative Let's join hands to protect the ecological environment and live a healthy life together. Environment/Join Hands/Co-create Declarative At this very moment, let's join the world in turning off the lights for one hour, taking action towards a more environmentally friendly lifestyle. Right now/Action/Committed to Declarative Protecting wildlife reflects humane care and values. Wildlife/Humane Care Pun Harnessing "nuclear" power to protect our green waters and mountains, supporting nature's wellbeing, we are taking action! Nature/Action Pun Use eco-friendly travel products on your journey. Be a responsible traveler and protect the ecosystem through small actions! Environment/Responsibility/Protection/Ecosystem/Action Pun Reduce a "carbon" footprint today, and receive nature's rewards tomorrow. Earth/ Nature/ Today Metaphor Together, let's love the Earth and live in harmony with nature. We are all children of the natural world. Earth/ Nature/Love Together 4.1.2 Theme analysis after the COVID-19 pandemic During the post-COVID-19 pandemic period, 246 blog posts and 34,787 comments were collected using Python 3.10. Following expert summarization and categorization, 25 clustered thematic sentences were identified, categorized into four types: rhetorical questions, metaphorical rhetoric, puns, and declarative expressions (Table 2). Declarative expressions accounted for 8 of the 25 themes, marking a significant decrease from 52.5% to 32%. Rhetorical questions made up 3 themes, reducing their share from 19% to 12%. Puns emerged as the dominant theme, accounting for 10 sentences and increasing from 19% to 40%. Finally, metaphorical rhetoric rose from 2 to 4 themes, raising its proportion to 16%. By analyzing temporal evolution, we identified new trends when comparing pre- and post-pandemic samples. A growing number of CSR advertisements are adopting various rhetorical theme types to engage with consumers. Notably, after the COVID-19 pandemic, the use of puns as a theme has significantly increased, becoming the preferred rhetorical strategy in many CSR advertising texts today. Through post-pandemic BERTopic theme clustering analysis, we discovered that CSR advertisements have shifted from focusing on "present" lifestyle discussions to emphasizing "future" considerations and actions. Additionally, more CSR advertisements are employing rhetorical strategies to refine their content, with the majority choosing puns as their primary rhetorical device. This trend underscores the growing attention and utilization of puns by CSR advertisements in recent years. Table 2: Themes After COVID-19 After COVID-19 Theme Types Examples of Thematic Sentences Keywords Pun "GOOD GOOD, Let's make the Earth less 'carbonated' and take action towards the future." Carbon Gas/Action/Future Pun Earth Day AI sci-fi blockbuster, Jane AI Earth in the distant future, what will happen to the Earth depends on our choices today, Jane AI Earth, man and nature live in harmony. AI/Cherish/Future/Earth/Nature Pun For the sake of Earth's tomorrow, I’ll stop “straw”ing! Starting now, let’s stop using straws for the Earth! Straws/Earth/Tomorrow Pun Green living, shared future, let the Earth “chick”en up with vitality. Future/Earth/Life Pun Less carbon "sighs," more TREES; join us in adding a touch of "green magic" to the Earth. Carbon footprint/Green initiatives/Earth Metaphor We are nature’s children, protecting nature is protecting our future selves. Nature/Children/Tomorrow/Protection Metaphor Earth, the mother of humanity, the cradle of life, is so beautiful and majestic, yet so kind and gentle. It is only this big and won't grow any bigger. If it is destroyed, we have nowhere else to go. Earth/Mother/Destruction/Cradle/Life Pun Daily dark chocolate, smartly protect the Earth, the future is shaped by our actions. Smart Protection/Earth/Future Metaphor Master the future, embrace electric power. Let’s charge the Earth together. Future/Earth/Recharge Question Without a low-carbon lifestyle, can our Earth still be healthy? Life/Low-carbon/Health/Earth Metaphor Every small "cold" of Mother Earth is a major disaster for all of humanity. Disaster/Humanity/Earth/Mother Declarative Cherish the Earth, we are always in action. Cherish/Earth/Action Pun Eating McDonald’s, don’t forget the small efforts we can make; let’s be environmentally conscious together towards the future. Small Efforts/Environmental Protection Declarative Be a friend of nature; take an hour to do something beneficial for the Earth. Let's protect a better future for nature, hand in hand. Nature/Friends/Hand in Hand Pun "Guarding" the Earth: Buying groceries? Why so much plastic? One eco-friendly bag is enough! No Plastic/Environmental/Guardianship/Earth Declarative Guardians of the national park use their hands to protect this beautiful yet fragile natural environment. National Park/Nature/Environment/Protection Question In the warm spring, around the time of awakening insects. Do you know how important the finless porpoise is to the health of the Yangtze River ecosystem? As the only remaining freshwater porpoise in the Yangtze, a first-class national protected wild animal, and a unique rare freshwater porpoise in China, its population is slowly recovering, but these creatures, with their healing smiles, still face severe survival threats. Ecosystem/Health/Wildlife Declarative Loving the Earth is not just about Earth Day; let’s adopt daily plastic reduction practices for a future-oriented lifestyle—reduce the use of plastic products. Eart/ Future/Plastic Reduction Declarative Sadly, since the pandemic, our Earth has faced a new form of pollution: mask pollution. Billions of discarded masks have been thrown away, entering the natural environment and becoming a nightmare for wildlife. We hope everyone will store masks properly and discard them at medical waste collection points in hospitals. Protecting the Earth is not just about slogans on Earth Day, but starts with small everyday actions. Pandemic/Pollution/Masks/Nightmare Question Can you find where the nature that represents positive change is? Let’s continue to protect biodiversity together and do something beneficial for the home we all depend on. Let’s gather, support each other, and be grateful for the beautiful home we have. Nature/Protection/Biodiversity/Survival Declarative Let’s join forces to keep life shining brightly in nature. With gentle strength, let’s preserve the beauty of nature so that we have no regrets tomorrow. Tomorrow/Life/Strength/Radiant/Nature Declarative Happy Earth Day to our only home. Home/Earth/Joy Declarative Landmarks around the world will turn off their lights; facing the future, we are stronger together. Lights Out/Future/Strength Pun The Low-Carbon Diary of the Great Sage Wins the Cycling Season. Choose green and low-carbon travel, protect the blue sky and white clouds, feel the gifts of nature, and make a promise to the future. Green Warrior/Low-carbo/ Future/Nature Pun Plastic reduction zone, please slow down on plastic use. Let’s speak up for environmental protection together. Reduce Plastic/Environmental Protection 4.2 Sentiment Analysis Following the completion of topic modeling, this study utilized SnowNLP to perform sentiment analysis on thematic sentences for each theme type, based on temporal evolution (Figure 2). As per the sentiment scoring settings of SnowNLP, the pun theme scored 0.71 pre-pandemic, rising to 0.75 post-pandemic, an increase of 0.04. This suggests that the pun theme has gained increasing popularity among consumers in recent years. The declarative expression theme scored 0.70 pre-pandemic but dropped significantly to 0.60 post-pandemic, indicating a substantial decline in sentiment. This implies that, following the onset of COVID-19, consumers have become weary and even resentful of CSR advertisement content lacking rhetorical strategies. Both metaphorical rhetoric and rhetorical question themes maintained neutral sentiment, with scores slightly increasing from 0.50 and 0.44 pre-pandemic to 0.52 and 0.50 post-pandemic, respectively. The minor increases in sentiment scores indicate that the pandemic did not significantly heighten consumer interest in these rhetorical themes. 4.3 Theme-Sentiment Evolution Analysis This study examines the rhetorical strategies in CSR advertising texts from both thematic and emotional perspectives. Based on this analysis, a Sankey diagram was created to illustrate the evolution of themes and sentiments, as depicted in Figure 3. The diagram illustrates the emotional flow of CSR advertising rhetorical themes across two distinct periods: pre-COVID-19 and post-COVID-19. Additionally, Tables 3 and 4 show the sentiment distribution proportions of themes for each period. Prior to the COVID-19 pandemic, as indicated in Table 3, declarative themes were the dominant mode of expression in CSR advertising texts. These themes exhibited a relatively high proportion of positive sentiment, suggesting that consumers were receptive to companies using this method to convey CSR messages. The proportions of neutral and negative sentiment further imply that consumers did not oppose this mode of expression. Previous research noted that pre-pandemic CSR advertising strategies frequently employed language emphasizing "now" and "the present," reflecting a shared effort by consumers and CSR advertisements to create an immediate consumption discourse environment. Declarative expressions, which convey information straightforwardly, effectively aligned with this immediate consumption context. Consequently, declarative expressions not only made up more than half of the thematic quantity compared to other themes during this period, but the high proportion of positive consumer sentiment also clearly demonstrated an interest in and preference for this mode of CSR advertising. As shown in Figure 4, the proportion of the use of puns and rhetorical questions as rhetorical themes during this period is equal. However, the emotional distribution among consumers differs significantly. Puns evoke the most positive emotions among consumers, whereas rhetorical questions evoke the highest proportion of negative emotions. These two rhetorical themes, with the same frequency but opposite consumer reactions, indicate that consumers prefer a more humorous and witty approach in green brand communication, which better aligns with their psychological preferences. Rhetorical questions primarily use techniques such as questioning to generate text content, often perceived as didactic and containing elements of preaching, such as "Do you want every drop of water to be clean? Then start saving water." Consumers tend to feel aversion and dissatisfaction towards such content. Metaphorical rhetoric themes were relatively underrepresented during the pre-COVID-19 period. Consumers predominantly expressed neutral sentiment towards these themes, which also had the lowest proportion of negative sentiment among the four theme types. Although metaphorical rhetoric was not frequently used as the primary textual strategy in CSR advertisements, its status as the theme with the least negative sentiment suggests it remains a preferred choice for CSR campaigns seeking to gauge consumer reactions. As shown in Table 4, following the COVID-19 pandemic, the proportion of declarative themes dropped significantly from 52.5% to 32%, marking a decrease of 39%. The proportion of positive sentiment similarly decreased from 0.6402 to 0.4195, dropping from the second to the third position. Meanwhile, negative sentiment rose from the third to the second position, whereas neutral sentiment remained relatively stable. These shifts in theme and sentiment proportions suggest that both CSR advertisers and consumers have become less receptive to straightforward declarative content after the pandemic. From the aggregated thematic keywords discussed earlier, it is evident that "future" and "tomorrow" have emerged as central contexts for themes post-pandemic, indicating a shift from immediate consumption to discussions about future lifestyles. This shift in contextual focus in CSR advertising has contributed to the decline of declarative expressions, as both advertisers and consumers now favor content that addresses the long-term impact of their actions. Consequently, consumers have grown increasingly indifferent to declarative themes. As illustrated in Figure 5, pun themes have emerged as the preferred choice for most CSR advertisements on social media, with their proportion increasing from 19% pre-pandemic to 40% post-pandemic. The positive sentiment proportion for pun themes consistently remains the highest among all theme types, maintaining a score above 0.7. This indicates that CSR advertisements are increasingly adopting puns for consumer engagement, with consumers appreciating this humorous and highly engaging rhetorical style. Examples include statements like, "For the sake of the Earth's tomorrow, I don't 'straw' it! From now on, for the Earth, no more straws!" and "GOOD GOOD, reduce 'carbon' emissions, act for the future." In the post-pandemic period, the proportion of metaphorical rhetoric themes increased from 9.5% to 16%. Although sentiment proportions did not change significantly compared to the pre-pandemic period, metaphorical rhetoric themes continued to show the lowest negative sentiment and the highest neutral sentiment among all themes. This indicates that some CSR advertisements are exploring consumer emotional attitudes and adjusting their emotional connections with consumers. The proportion of rhetorical question themes has decreased from 19% to 12%, making it the least used theme type in CSR advertisements. Not only have CSR advertisements gradually abandoned this theme, but consumers continue to show aversion towards it. The proportion of negative sentiment has not improved despite the shift towards "future" oriented content, making it still the most disliked rhetorical theme among consumers. The distribution of rhetorical types after the COVID-19 pandemic has shown significant changes compared to before the pandemic, as depicted in Figures 4 and 5. In positive emotions, the proportion of declarative has decreased, while the use of pun and metaphor has increased. This shift may reflect a tendency during the pandemic for people to employ humorous and symbolic expressions to alleviate stress. Conversely, in negative emotions, there has been a notable increase in the proportion of questions, possibly related to the uncertainties, challenges, and difficulties brought about by the pandemic, as people express doubts and concerns through questioning. Furthermore, examining rhetorical changes across emotional categories reveals varied impacts of the COVID-19 on rhetorical proportions. In positive emotions, there appears to be greater diversity in rhetorical types, with an increase not only in direct declarative expressions but also in the use of pun and metaphor to enrich emotional expression. In neutral emotions, changes in rhetorical types have been relatively stable, yet there has been an increase in the use of metaphor, possibly reflecting a heightened preference for symbolic and metaphorical expressions during the pandemic. In negative emotions, the increase in the proportion of questions is particularly pronounced, further emphasizing the negative impact of the pandemic on people's lives and mental health. In summary, the COVID-19 pandemic has significantly influenced the distribution of rhetorical types. The increase in pun and metaphor in positive emotions reflects a trend towards seeking humor and symbolic expression during the pandemic. The notable increase in questions in negative emotions highlights the pandemic's adverse effects on people's lives and mental health. These changes not only reflect the pandemic's impact on rhetorical usage but also reveal shifts in people's psychological responses and adaptation processes during this challenging time. Table 3 Theme Sentiment Distribution Before the COVID-19 Pandemic Theme Type Proportion of Theme Quantity Proportion of Positive Sentiment Proportion of Neutral Sentiment Proportion of Negative Sentiment Declarative 52.50% 0.6402 0.2682 0.0916 Pun 19% 0.7133 0.1515 0.135 Question 19% 0.175 0.5225 0.3025 Metaphor 9.50% 0.124 0.7895 0.0865 Table 4 Theme Sentiment Distribution After the COVID-19 Pandemic Theme Type Proportion of Theme Quantity Proportion of Positive Sentiment Proportion of Neutral Sentiment Proportion of Negative Sentiment Declarative 32% 0.4195 0.4169 0.1636 Pun 40% 0.7082 0.2237 0.0681 Question 12% 0.3037 0.4537 0.2427 Metaphor 16% 0.1788 0.7725 0.0488 5 Discussion of Research Findings In terms of text theme mining, this paper analyzes the microblog texts of the pre-New Crown epidemic and post-New Crown epidemic phases, and identifies the shift of CSR advertisements from declarative to punning rhetorical themes in generating textual content, and according to the aggregated data on theme word extraction, it can be found that the expression of textual content has changed from the present to the future, and this change also proves that the major public health emergency has made CSRs focus more on “consumers' present behavior will affect the future of society” when constructing advertisements. The change from “present” to “future” in creating the context of text content is also evidence that the emergence of a major public health event has made CSRs focus more on “how consumer behavior in the present will affect society in the future” in constructing the content of advertisements. In terms of sentiment analysis, consumers' sentiment scores for CSR advertisements are in the distribution range of neutral sentiment share, and from the scores of the pre-stage and post-stage of the New Crown Epidemic, the pre-stage is 0.59, and the post-stage score is 0.60, which can indicate that after the New Crown Epidemic, consumers pay more attention to and like the CSR advertisements and it also indicates that the suddenness of the major public health event from the side makes consumers have positive emotions towards CSR advertisements and the sustainable behaviors they advocate. In terms of topic-emotion evolution analysis, before the Xin Guan epidemic, most CSR advertisements preferred to use statement expressions, which are simple, clear in purpose, and do not contain rhetorical content, for marketing communication on social platforms, and consumers have more positive emotional expressions for them, but with the changes in the social environment, the original topic context and type of topic expression gradually changed when CSR advertisements generated text content on social platforms. However, with the change of social environment, CSR advertisements have gradually changed the original topic context and theme expression type when generating text content on social platforms, and the number of CSR advertisements containing declarative themes has declined, and has been replaced by punning rhetorical themes as the main rhetorical mode of CSR advertisements in social media marketing communication. Consumers are more likely to accept CSR advertisements that use humorous semantics and attention-grabbing language styles than statements. Therefore, punning rhetoric has become the main way of marketing communication of CSR ads to consumers in the post epidemic era. 6 Research Significance and Shortcomings This study utilized the BERTopic model to conduct theme clustering analysis of the rhetorical content in CSR advertisements on Chinese social media platforms and employed SnowNLP for sentiment scoring. Based on previous research findings, this study first demonstrated that the rhetorical themes in CSR advertisement content shifted before and after the COVID-19 pandemic. Puns have become the commonly used rhetorical strategy in CSR ads, and consumers maintain a positive emotional attitude towards this theme type. This finding fills the gap in CSR research concerning the relationship between CSR advertising and consumer emotions in the context of Chinese language, providing a basis for further study. Secondly, due to the COVID-19 pandemic, the semantic context on Chinese social media shifted from focusing on the "present" to considering the "future." This shift led to a significant increase in the use of rhetorical strategies in CSR advertisements, demonstrating that unexpected events can significantly impact the relationship between online semantic contexts and CSR advertising communication. This in-depth study of CSR advertising at different stages has established a relationship between CSR advertisements and changes in consumer emotions, enriching related research. However, several limitations remain: 1. This study only compared data samples from 2017-2019 and 2021-2023, without conducting long-term tracking. Future research should consider changes over a longer time scale. 2. This study used only the BERTopic clustering model and SnowNLP sentiment analysis, both suitable for short text processing. Since CSR advertising involves more than just Weibo posts, future studies could incorporate other methods for comparative analysis. Declarations Author Contribution J.W. conceptualized the study, conducted the data analysis, and drafted the initial version of the manuscript. X.S. participated in the study design, assisted with data interpretation, and critically revised the manuscript. W.Z., as the corresponding author, contributed to the writing and editing of the manuscript. Y.J. was responsible for data cleansing and collation, while Y.L. verified and proofread the data. All authors reviewed and approved the final manuscript.Informed consent were obtained from all the paticipants. Acknowledgement The authors would like to express their gratitude to the National Natural Science Foundation of China ( No. 72102172) for their financial support of this research. Data Availability Data Availability StatementThe data that support the findings of this study are not publicly available as they are part of an ongoing research project. For any inquiries regarding the data, please contact the corresponding author（Dr. wei ZHANG / [email protected] ）. The data will be provided if necessary for review purposes or specific requests. References Angelov, D. (2020). Top2Vec: Distributed Representations of Topics (arXiv:2008.09470). arXiv. http://arxiv.org/abs/2008.09470 Bergkvist, L., & Zhou, K. Q. (2019). Cause-related marketing persuasion research: An integrated framework and directions for further research. International Journal of Advertising, 38(1), 5–25. https://doi.org/10.1080/02650487.2018.1452397 Cai, M. (2021). Natural language processing for urban research: A systematic review. Heliyon, 7(3), e06322. https://doi.org/10.1016/j.heliyon.2021.e06322 Carroll, A. B. (2021). Corporate social responsibility (CSR) and the COVID-19 pandemic: Organizational and managerial implications. Journal of Strategy and Management, 14(3), 315–330. https://doi.org/10.1108/JSMA-07-2021-0145 Chang, C. (2012). The effectiveness of advertising that leverages sponsorship and cause-related marketing: A contingency model. International Journal of Advertising, 31(2), 317–338. https://doi.org/10.2501/IJA-31-2-317-337 Chen, C., Chen, J., & Shi, C. (2018). Research on Credit Evaluation Model of Online Store Based on SnowNLP. E3S Web of Conferences, 53, 03039. https://doi.org/10.1051/e3sconf/20185303039 Chernev, A., & Blair, S. (2015). Doing Well by Doing Good: The Benevolent Halo of Corporate Social Responsibility. Journal of Consumer Research, 41(6), 1412–1425. https://doi.org/10.1086/680089 Choi, C. S., Cho, Y.-N., Ko, E., Kim, S. J., Kim, K. H., & Sarkees, M. E. (2019). Corporate sustainability efforts and e-WOM intentions in social platforms. International Journal of Advertising, 38(8), 1224–1239. https://doi.org/10.1080/02650487.2019.1613858 Choi, C.-W. (2022). Increasing company-cause fit: The effects of a relational ad message and consumers’ cause involvement on attitude toward the CSR activity. International Journal of Advertising, 41(2), 333–353. https://doi.org/10.1080/02650487.2020.1822060 Diehl, S., Terlutter, R., & Mueller, B. (2016). Doing good matters to consumers: The effectiveness of humane-oriented CSR appeals in cross-cultural standardized advertising campaigns. International Journal of Advertising, 35(4), 730–757. https://doi.org/10.1080/02650487.2015.1077606 Ditlev-Simonsen, C., & Wenstøp, S. (n.d.). Companies’ ethical commitment: An analysis of the rhetoric in CSR reports. Eberle, D., Berens, G., & Li, T. (2013). The Impact of Interactive Corporate Social Responsibility Communication on Corporate Reputation. Journal of Business Ethics, 118(4), 731–746. https://doi.org/10.1007/s10551-013-1957-y Egger, R., & Yu, J. (2022). A Topic Modeling Comparison Between LDA, NMF, Top2Vec, and BERTopic to Demystify Twitter Posts. Frontiers in Sociology, 7, 886498. https://doi.org/10.3389/fsoc.2022.886498 Feldner, S. B. (2014). How Corporations Manage Industry and Consumer Expectations via the CSR Report. Public Relations Journal, 8(3). Friestad, M., & Wright, P. (1994). The Persuasion Knowledge Model: How People Cope with Persuasion Attempts. Journal of Consumer Research, 21(1), 1–31. https://doi.org/10.1086/209380 Gao, Y. (2011). CSR in an emerging country: A content analysis of CSR reports of listed companies. Baltic Journal of Management, 6(2), 263–291. https://doi.org/10.1108/17465261111131848 Grootendorst, M. (2022). BERTopic: Neural topic modeling with a class-based TF-IDF procedure (arXiv:2203.05794). arXiv. http://arxiv.org/abs/2203.05794 Hartmann, P., Marcos, A., & Apaolaza, V. (2023). Past, present, and future of research on corporate social responsibility advertising. International Journal of Advertising, 42(1), 87–95. https://doi.org/10.1080/02650487.2022.2121902 Hu, N. (2022). Sentiment Analysis of Texts on Public Health Emergencies Based on Social Media Data Mining. Computational and Mathematical Methods in Medicine, 2022, 1–7. https://doi.org/10.1155/2022/3964473 Huang, C., Zhuang, S., Li, Z., & Gao, J. (2022). Creating a Sincere Sustainable Brand: The Application of Aristotle’s Rhetorical Theory to Green Brand Storytelling. Frontiers in Psychology, 13, 897281. https://doi.org/10.3389/fpsyg.2022.897281 Jing Wen, T., Kim, E., Wu, L., & Dodoo, N. A. (2020). Activating persuasion knowledge in native advertising: The influence of cognitive load and disclosure language. International Journal of Advertising, 39(1), 74–93. https://doi.org/10.1080/02650487.2019.1585649 Lee, S. Y., & Carroll, C. E. (2011). The Emergence, Variation, and Evolution of Corporate Social Responsibility in the Public Sphere, 1980–2004: The Exposure of Firms to Public Debate. Journal of Business Ethics, 104(1), 115–131. https://doi.org/10.1007/s10551-011-0893-y Li, S. (Kevin), Wang, L., Huang, J., Gopal, R., & Lin, Z. (2024). Sizzle or Fizzle? Supply and Consumption Dynamics of Home-Cooked Food on Sharing Platform. Production and Operations Management, 33(1), 11–31. https://doi.org/10.1177/10591478231224954 Liu, C., Chen, S., & Shao, Q. (2019). Do CEO Rhetorical Strategies Affect Corporate Social Performance? Evidence from China. Sustainability, 11(18), 4907. https://doi.org/10.3390/su11184907 Lloyd-Smith, P., & An, H. (2019). Are corporate social responsibility and advertising complements or substitutes in producing firm reputation? Applied Economics, 51(21), 2275–2288. https://doi.org/10.1080/00036846.2018.1540858 Lowrey, T. M. (2002). Does language matter? Psychology & Marketing, 19(7–8), 569–572. https://doi.org/10.1002/mar.10025 Luangrath, A. W., Peck, J., & Barger, V. A. (2017). Textual paralanguage and its implications for marketing communications. Journal of Consumer Psychology, 27(1), 98–107. https://doi.org/10.1016/j.jcps.2016.05.002 Marais, M. (2012). CEO rhetorical strategies for corporate social responsibility (CSR). Society and Business Review, 7(3), 223–243. https://doi.org/10.1108/17465681211271314 McInnes, L., & Healy, J. (2017). Accelerated Hierarchical Density Based Clustering. 2017 IEEE International Conference on Data Mining Workshops (ICDMW), 33–42. https://doi.org/10.1109/ICDMW.2017.12 McInnes, L., Healy, J., Saul, N., & Großberger, L. (2018). UMAP: Uniform Manifold Approximation and Projection. Journal of Open Source Software, 3(29), 861. https://doi.org/10.21105/joss.00861 Panagiotopoulos, I. (2021). Novel CSR & novel coronavirus: Corporate social responsibility inside the frame of coronavirus pandemic in Greece. International Journal of Corporate Social Responsibility, 6(1), 10. https://doi.org/10.1186/s40991-021-00065-7 Pogacar, R., Shrum, L. J., & Lowrey, T. M. (2018). The Effects of Linguistic Devices on Consumer Information Processing and Persuasion: A Language Complexity × Processing Mode Framework. Journal of Consumer Psychology, 28(4), 689–711. https://doi.org/10.1002/jcpy.1052 Rahman, M., Rodríguez-Serrano, M. Á., & Lambkin, M. (2017). Corporate Social Responsibility And Marketing Performance: The Moderating Role Of Advertising Intensity. Journal of Advertising Research, 57(4), 368–378. https://doi.org/10.2501/JAR-2017-047 Reimers, N., & Gurevych, I. (2019). Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks (arXiv:1908.10084). arXiv. http://arxiv.org/abs/1908.10084 Sánchez‐Franco, M. J., & Rey‐Moreno, M. (2022). Do travelers’ reviews depend on the destination? An analysis in coastal and urban peer‐to‐peer lodgings. Psychology & Marketing, 39(2), 441–459. https://doi.org/10.1002/mar.21608 Schouten, A. P., Janssen, L., & Verspaget, M. (2020). Celebrity vs. Influencer endorsements in advertising: The role of identification, credibility, and Product-Endorser fit. International Journal of Advertising, 39(2), 258–281. https://doi.org/10.1080/02650487.2019.1634898 Taboada, M., & Mann, W. C. (2006). Applications of Rhetorical Structure Theory. Discourse Studies, 8(4), 567–588. https://doi.org/10.1177/1461445606064836 Tomaselli, G., & Melia, M. (2014). THE ROLE OF INTERACTIVE TECHNOLOGIES FOR CSR COMMUNICATION. 8. Wang, C. L. (2021). New frontiers and future directions in interactive marketing: Inaugural Editorial. Journal of Research in Interactive Marketing, 15(1), 1–9. https://doi.org/10.1108/JRIM-03-2021-270 Wang, H.-J. (2019). Green city branding: Perceptions of multiple stakeholders. Journal of Product & Brand Management, 28(3), 376–390. https://doi.org/10.1108/JPBM-07-2018-1933 Xie, Q., & Wang, T. (Grace). (2022). Promoting corporate social responsibility message in COVID-19 advertising: How threat persuasion affects consumer responses to altruistic versus strategic CSR. Journal of Business Research, 148, 315–324. https://doi.org/10.1016/j.jbusres.2022.04.073 Xu, J., Guo, D., Xu, J., & Luo, C. (2023). How Do Multiple Actors Conduct Science Communication About Omicron on Weibo: A Mixed-Method Study. Media and Communication, 11(1), 306–322. https://doi.org/10.17645/mac.v11i1.6122 Yu, S., Eisenman, D., & Han, Z. (2021). Temporal Dynamics of Public Emotions During the COVID-19 Pandemic at the Epicenter of the Outbreak: Sentiment Analysis of Weibo Posts From Wuhan. Journal of Medical Internet Research, 23(3), e27078. https://doi.org/10.2196/27078 Zhang, Z., Feng, G., Xu, J., Zhang, Y., Li, J., Huang, J., Akinwunmi, B., Zhang, C. J. P., & Ming, W. (2021). The Impact of Public Health Events on COVID-19 Vaccine Hesitancy on Chinese Social Media: National Infoveillance Study. JMIR Public Health and Surveillance, 7(11), e32936. https://doi.org/10.2196/32936 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-5697919","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":398817580,"identity":"f8114d48-72ed-4a56-afb2-827f43138b41","order_by":0,"name":"jianguo wang","email":"","orcid":"","institution":"Wuhan University of Technology","correspondingAuthor":false,"prefix":"","firstName":"jianguo","middleName":"","lastName":"wang","suffix":""},{"id":398817581,"identity":"7f941b76-a5aa-4ee5-adbb-10ec3ab8a6d9","order_by":1,"name":"xixiang sun","email":"","orcid":"","institution":"Wuhan University of Technology","correspondingAuthor":false,"prefix":"","firstName":"xixiang","middleName":"","lastName":"sun","suffix":""},{"id":398817582,"identity":"0bb4aee7-8571-4748-8ccb-4a94d526c19e","order_by":2,"name":"wei zhang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA9klEQVRIiWNgGAWjYDACCQjFw97AfIAZwk4gUgvPAbYEoBYD4rUw8BzgMSBOi/zs5mcPv7bZyPBI5HyTLmz7w8DPnmPA8HMHbi2Mc46ZG8u2pfHwSORuk57ZZsAg2fPGgLH3DG4tzBIJZtKS2w7z2AO13OYFajG4kWPAzNiGWwubRPo3oJb/QFtynoG12BPSAlRpJvlx2wGQFjaILRIEtEhI5JRJM/5L5uHheWb+m+ecMY/EmWcFB3vxaJGfkb5N8scZO3se9uTHxjxlcnL87ckbH/zEowUcBDwgUiAB4lIQcQC/BmBA/wCR/ATVjYJRMApGwUgFACnDR9VVlapJAAAAAElFTkSuQmCC","orcid":"","institution":"Wuhan University of Technology","correspondingAuthor":true,"prefix":"","firstName":"wei","middleName":"","lastName":"zhang","suffix":""},{"id":398817583,"identity":"932cf26a-c130-4725-88e2-a682e7f90674","order_by":3,"name":"yige jia","email":"","orcid":"","institution":"Wuhan University of Technology","correspondingAuthor":false,"prefix":"","firstName":"yige","middleName":"","lastName":"jia","suffix":""},{"id":398817584,"identity":"7fa7c172-dba0-4c72-9cce-d8e731561264","order_by":4,"name":"you li","email":"","orcid":"","institution":"Wuhan University of Technology","correspondingAuthor":false,"prefix":"","firstName":"you","middleName":"","lastName":"li","suffix":""}],"badges":[],"createdAt":"2024-12-23 08:38:23","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5697919/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5697919/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":73399280,"identity":"3517d096-3dde-4d77-88a3-b7333239df57","added_by":"auto","created_at":"2025-01-09 14:15:57","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":156525,"visible":true,"origin":"","legend":"\u003cp\u003eResearch framework\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-5697919/v1/d8510c4ca32d55f7e6fe9673.png"},{"id":73398163,"identity":"432083ff-727a-4967-aa3e-4c8f7c9bd2f2","added_by":"auto","created_at":"2025-01-09 14:07:57","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":16442,"visible":true,"origin":"","legend":"\u003cp\u003eSentiment Scores for Different Theme Types\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-5697919/v1/3dd19f13ec6590371c0230d0.png"},{"id":73399286,"identity":"e32d5435-634e-4376-8b6e-dfad2ecfa32e","added_by":"auto","created_at":"2025-01-09 14:15:57","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":17032415,"visible":true,"origin":"","legend":"\u003cp\u003eDynamic Evolution of Themes and Sentiments in CSR Advertising Text Rhetorical Strategies\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-5697919/v1/8c32b4117734779ac6d59587.png"},{"id":73398165,"identity":"5b308a17-fa39-4ffa-ae41-c28193b76df5","added_by":"auto","created_at":"2025-01-09 14:07:57","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":7899,"visible":true,"origin":"","legend":"\u003cp\u003eStacked Graph of Emotional Percentage under Different Rhetoric before COVID-19\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-5697919/v1/af2a7d866133b9215e0b992d.png"},{"id":73398171,"identity":"74bdd389-cce5-4166-b1c0-fa22b6a2d838","added_by":"auto","created_at":"2025-01-09 14:07:57","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":9338,"visible":true,"origin":"","legend":"\u003cp\u003eStacked Graph of Emotional Percentage under Different Rhetoric after COVID-19\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-5697919/v1/473a7644cada52a8e191fc68.png"},{"id":79452957,"identity":"4b68f352-02e9-489f-a696-848e932c940d","added_by":"auto","created_at":"2025-03-28 15:17:20","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":22962243,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5697919/v1/316619cc-c7c6-4154-8b17-3b6c2560fd0c.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"The Evolution of Consumer Emotions in Response to Rhetorical Strategies in Corporate Social Responsibility Advertising: Evidence from Chinese Social Networks Before and After the COVID-19 Pandemic","fulltext":[{"header":"0 Introduction","content":"\u003cp\u003eDriven by technological advancements, commercial companies are exploring innovative methods to create value by connecting with consumers and enhancing engagement and interaction (Wang,2021). As information technology evolves and social media rises, interactive marketing has become crucial for companies involved in corporate social responsibility (CSR) efforts. Higher perceived interactivity in CSR communications can enhance information credibility and strengthen consumer corporate identification, improving reputation and increasing willingness to share positive word-of-mouth(Eberle et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). Interactive technologies, like websites and social media platforms, make CSR communication more engaging and foster dialogue between companies and consumers (Tomaselli \u0026amp; Melia, \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). Additionally, CSR significantly impacts corporate performance when communicated through advertising (Rahman et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Advertising intensity positively moderates the relationship between CSR activities and market share(Rahman et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Both advertising and CSR are essential for companies to build reputation, enhance brand authenticity, and boost credibility (Lloyd-Smith \u0026amp; An, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Since its inception, research on corporate social responsibility has been closely linked with brand building, corporate advertising, and consumer emotions (Hartmann et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2023\u003c/span\u003e)。\u003c/p\u003e \u003cp\u003eThe global COVID-19 pandemic has significantly impacted the brand image of companies that have long positioned themselves as green, environmentally friendly, and sustainable. Additionally, the pandemic has shaken consumer confidence in CSR initiatives (Panagiotopoulos, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). This global health crisis is testing companies' commitment to CSR, with evidence suggesting that many are reevaluating and adjusting their CSR strategies to better adapt to the crisis and meet changing public expectations(Carroll, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). In response to pandemic challenges, many companies are incorporating more altruistic messaging in CSR advertising to rebuild trust and strengthen the consumer relationship with CSR initiatives(Xie \u0026amp; Wang, 2022). As we move into the post-COVID-19 era, it is increasingly important to explore how companies can use CSR advertising effectively to influence consumer emotions and restore confidence in CSR practices. This study aims to examine the relationship between CSR advertising and consumer emotions by investigating strategic approaches to the processing of textual information in CSR advertisements.\u003c/p\u003e \u003cp\u003eOur study aims to investigate the emotional connections between rhetorical strategies used in CSR advertisements and consumer reactions by applying cluster analysis to CSR ads gathered from Chinese social networking platforms. This research seeks to identify specific rhetorical techniques that, when used in creating structured and curated textual content, effectively elicit positive emotional responses from consumers. By understanding these strategies, we aim to reveal how CSR advertising can foster closer, more intimate relationships between consumers and CSR initiatives.\u003c/p\u003e \u003cp\u003eThe structure of this study is as follows: First, a literature review is conducted, analyzing previous research on corporate social responsibility, advertising communication, and textual rhetorical strategies. Next, the methodology section outlines the data collection and cleaning processes. The study employs clustering analysis using the BERTopic model(Grootendorst, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), which is based on the Top2Vec framework(Angelov, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Subsequently, SnowNLP is applied to perform sentiment analysis on the rhetorical strategies identified within the clusters(Cai, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). The paper concludes with a discussion of the findings and provides recommendations for future research.\u003c/p\u003e"},{"header":"1 Literature Review","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e1.1 Current State of CSR Advertising Research\u003c/h2\u003e \u003cp\u003eCSR advertising is a vital area of study, examining how companies use advertising to communicate CSR activities to consumers and build relationships. Research has explored the dual objectives of CSR advertising: enhancing corporate reputation and positively influencing consumer attitudes. A significant focus of CSR advertising research is its impact on corporate reputation. Studies show that effective CSR advertising can lead to favorable consumer perceptions of a company\u0026rsquo;s image, enhancing brand equity (Chang, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Chernev \u0026amp; Blair, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). Consumers generally respond positively to CSR ads that align with their values, highlighting the importance of strategic message alignment in CSR campaigns (Diehl et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). Another key research area is stakeholder reactions, particularly consumer and employee responses to CSR advertising. Literature suggests that positive consumer responses often depend on the perceived authenticity and alignment between a company's CSR efforts and its brand identity (C.-W. Choi, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Consumers are more likely to trust and support companies whose CSR activities appear genuinely committed to social good, rather than being self-serving (Bergkvist \u0026amp; Zhou, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe rise of digital platforms has dramatically reshaped CSR advertising strategies. Research highlights the growing influence of social media and electronic word-of-mouth (eWOM) in CSR communication. Social media is a powerful tool for disseminating CSR messages, facilitating consumer engagement, and co-creating content, thereby amplifying the reach and impact of CSR initiatives (Schouten et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Studies also emphasize the role of influencers in CSR advertising, noting that the trustworthiness and perceived authenticity of influencers shape consumer attitudes toward CSR messages(C. S. Choi et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Authenticity remains a crucial theme in CSR advertising. According to the Persuasion Knowledge Model, consumers' ability to recognize and evaluate the authenticity of CSR efforts significantly influences their reactions (Friestad \u0026amp; Wright, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e1994\u003c/span\u003e). When consumers perceive CSR advertisements as insincere or purely profit-driven, skepticism may arise, potentially damaging the brand's reputation (Jing Wen et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).Thus, maintaining authenticity in CSR messaging is essential to prevent consumer backlash.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003e1.2 Current Research on Textual Rhetorical Strategies\u003c/h3\u003e\n\u003cp\u003eLanguage is a crucial tool in marketing communication. As early as 1923, American communication pioneer Claude C. Hopkins noted in \u003cem\u003eScientific Advertising\u003c/em\u003e that copywriting rhetoric, sentence structure arrangement, and headline wording strongly appeal to consumers. In today's interactive marketing environment, understanding consumers' linguistic psychology helps green marketers communicate more effectively and reach consumers' minds swiftly., For example, green brands like McDonald's and Starbucks have captured consumer attention, sparked discussion, and encouraged active sharing through eco-friendly content on Chinese social media platforms. The use of language in marketing communication has been a long-standing area of interest. Numerous studies have applied psycholinguistic paradigms and methods to marketing contexts(Lowrey, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2002\u003c/span\u003e).These studies show that language elements in marketing\u0026mdash;from brand naming and slogans to copywriting rhetoric, advertising voiceovers, and product descriptions\u0026mdash;significantly impact consumer memory, attitudes, and behavior(Pogacar et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2018\u003c/span\u003e) .In marketing, textual language functions to transmit information, convey values, explain, and persuade(Luangrath et al., \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2017\u003c/span\u003e), Therefore, existing research primarily focuses on how different rhetorical uses in text affect consumer emotional responses.\u003c/p\u003e \u003cp\u003eRecent research shows a close relationship between textual rhetorical strategies and CSR..CEOs use various rhetorical methods in CSR communication, such as ethos, pathos, and logos, where pathos and logos positively impact corporate social performance(Liu et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Three primary categories of rhetoric are identified: value rhetoric for moral legitimacy, normative rhetoric for cognitive legitimacy, and instrumental rhetoric for pragmatic legitimacy (Marais, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). An analysis of CSR reports shows that ethos is the most commonly used moral perspective, followed by benefit and agency(Ditlev-Simonsen \u0026amp; Wenst\u0026oslash;p, n.d.). Companies employ these rhetorical strategies in CSR reports to manage stakeholder expectations and legitimize their actions(Feldner, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). The most prevalent pattern links agency and benefit perspectives, suggesting that the pursuit of profit is justified by agency. These findings highlight the importance of rhetorical strategies in CSR communication and their impact on stakeholder perceptions. Current academic research on the relationship between CSR and rhetorical strategies predominantly centers on benefit-related perspectives. However, limited research has explored the connection between the specific wording used in texts and CSR. Therefore, this study seeks to uncover the relationship between textual rhetoric in CSR advertisements and consumer emotions.\u003c/p\u003e\n\u003ch3\u003e1.3 State of the Art in Thematic Modeling Research\u003c/h3\u003e\n\u003cp\u003eLatent Dirichlet Allocation (LDA) is a generative statistical model commonly used for topic modeling to discover abstract topics in document collections (Blei, Ng, \u0026amp; Jordan, 2003). LDA assumes each document is a mixture of a few topics, with each word attributable to one of these topics. Since its introduction, LDA has become a standard tool for text analysis, valued for its robustness in handling large datasets and its capacity to uncover hidden thematic structures. LDA has been widely applied in CSR research to analyze and understand various dimensions of CSR activities and their communication. Several key studies have demonstrated LDA's effectiveness in analyzing CSR-related content. For instance, Lee and Carroll (Lee \u0026amp; Carroll, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2011\u003c/span\u003e) used LDA to examine CSR reports from Fortune Global 500 companies, identifying key themes such as environmental management, employee welfare, and ethical governance. Another study by Gao, Yu, and Lee(Gao, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2011\u003c/span\u003e) applied LDA to analyze CSR disclosures in China, highlighting how CSR themes evolved over time. However, LDA, which relies on the bag-of-words approach, often fails to capture semantic relationships between words.. Because this method ignores the context of word appearance, it cannot accurately represent document content. To address these limitations, Maarten Grootendorst introduced the BERTopic model in 2020, which better captures the intricate relationships within text. Research in topic modeling has seen continuous advancements tailored to different textual characteristics. Traditional LDA models often face data sparsity issues, particularly in short texts, where sparse data limits reliable extraction of word co-occurrence patterns. Therefore, this study utilizes the BERTopic model to overcome limitations associated with traditional LDA topic modeling. The BERTopic model offers enhanced semantic interpretability, enabling it to better understand semantic connections and contextual information within text data, thereby more precisely capturing the underlying meaning of topics.\u003c/p\u003e \u003cp\u003eSentiment analysis is a widely used technique in social media data analysis, with methods broadly categorized into dictionary-based and machine learning-based approaches. Dictionary-based sentiment analysis uses pre-constructed sentiment lexicons to analyze emotions in text, such as the Hownet Sentiment Lexicon developed by China National Knowledge Infrastructure (CNKI). In contrast, machine learning-based sentiment analysis involves training models to recognize and classify emotions from data, often offering higher accuracy. Examples include using Recurrent Neural Networks (RNNs) and Convolutional Neural Networks (CNNs) for sentiment analysis. Some studies have used pre-packaged third-party sentiment analysis libraries to conduct sentiment analysis. Yu et al. utilized SnowNLP to detect emotions on Chinese social media platforms (Weibo) during the COVID-19 pandemic(Yu et al., \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) .Chinese scholars, Zhang et al., used SnowNLP to conduct sentiment analysis and monitor concerns about COVID-19 vaccines on Weibo(Zhang et al., \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2021\u003c/span\u003e).Researchers used SnowNLP and descriptive statistics to analyze social networks, demonstrating interactions among different participants, and used social media analysis to assess the impact of Weibo on public emotions(Xu et al., \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2023\u003c/span\u003e)。Chinese scholar Hu(Hu, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) utilized SnowNLP to analyze the emotional trends of online public opinion during major epidemics. Sentiment analysis, as an important tool in text analysis, requires careful selection based on different text analysis scenarios. Different sentiment analysis tools have their own strengths and applicable scenarios. In the specific context of social media sentiment analysis, the widespread use of SnowNLP has demonstrated its superior performance. Therefore, this study selects SnowNLP as the tool for sentiment analysis.\u003c/p\u003e"},{"header":"2 Research Objective","content":"\u003cp\u003eThis study aims to address the following primary research objectives (RO): By analyzing semantic dissemination texts on social media platforms, we seek to explore the relationship between textual rhetorical strategies used in CSR advertisements and consumer emotions, thereby establishing large-scale textual comparisons in this expanding field of research. Therefore, this study specifically proposes the following research objectives:\u003c/p\u003e \u003cp\u003eRO.1: By preprocessing (cleaning data) and applying data mining techniques to a large volume of CSR advertisement texts from social media platforms, this study aims to compare the types of rhetorical themes used in CSR advertising content before and after the COVID-19 pandemic.\u003c/p\u003e \u003cp\u003eRO.2. By clustering CSR advertisement content from before and after the COVID-19 pandemic and extracting thematic keywords to identify their moral context, this study explores the relationship between CSR advertising themes and their moral context.\u003c/p\u003e \u003cp\u003eRO.3. To explore the impact of rhetorical strategies used in CSR advertisements on consumers, this study seeks to clarify the relationship between the rhetorical themes of CSR advertising and consumer emotions.\u003c/p\u003e \u003cp\u003eOur research reveals the following: (1) it overcomes the limitations of probabilistic generative topic models; (2) it expands upon previous research on text processing and semantic themes; and (3) it applies exploratory methods using large-scale data sources. Therefore, compared to previous studies that relied on limited sample data, this research offers greater reliability and accuracy.\u003c/p\u003e"},{"header":"3 Research Design","content":"\u003cp\u003eAs illustrated in Figure 1, this study\u0026apos;s research framework for theme-emotion integration in CSR advertising rhetorical strategies starts with using Python 3.10 to collect target textual data from the Weibo platform, followed by data preprocessing and classification. Next, the BERTopic model is used to identify relevant thematic content in CSR advertising rhetorical strategies across different periods, before and after the COVID-19 pandemic. Additionally, SnowNLP analyzes the emotional evolution of consumer comments, revealing the emotional impact of CSR advertising rhetorical strategies on consumers.\u003c/p\u003e\n\u003ch2\u003e3.1 Data Collection\u003c/h2\u003e\n\u003cp\u003eThis study examines the rhetorical content of CSR advertisements. Recognizing that environmental activities can influence consumers\u0026apos; identity associations(H.-J.Wang, 2019) , we selected topics like #World Water Day (March 22), #World Earth Day (April 22), and #World Environment Day (June 5) to emphasize CSR identity. We then collected related texts from the Weibo platform. Data collected included user IDs, posting times, post content, number of likes, and comment texts. Our analysis was restricted to Simplified Chinese text data to ensure consistency. The study utilized Python 3.10 data packages. To ensure scientific rigor and data validity, we adopted S\u0026aacute;nchez-Franco and Rey-Moreno\u0026apos;s(S\u0026aacute;nchez‐Franco \u0026amp; Rey‐Moreno, 2022) framework, dividing data collection into two periods: pre-COVID-19 (March 22, 2017 - June 5, 2019) and post-COVID-19 (March 22, 2021 - June 5, 2023). As a result, the dataset comprised 834 microblog posts and 109,474 comments.\u003c/p\u003e\n\u003ch2\u003e3.2 Data Cleaning\u003c/h2\u003e\n\u003cp\u003eDuring the data cleaning phase, many texts contained terms related to specific interactive marketing activities, such as \u0026quot;World Earth Day,\u0026quot; \u0026quot;World Water Day,\u0026quot; and \u0026quot;World Environment Day.\u0026quot;. To enhance segmentation accuracy, we built a custom dictionary. While the BERTopic model imposes no restrictions on sample data, based on S\u0026aacute;nchez‐Franco and Rey‐Moreno（2022) , we aimed to better extract relevant text features, reduce dataset dimensions, and ensure data consistency.(1)We removed duplicate data;(2)Excluded non-brand-generated content;(3)Deleted common stop words to filter out frequent terms and customize proper nouns;(4)Employed jieba for text segmentation;(5)Eliminated noise text;(6)Tokenized symbols and emojis in comments. In the BERTopic model, each sentence is considered an independent text unit. Therefore, we applied sentence-level processing to more effectively capture subtle differences. The final dataset consisted of 526 pieces of green brand-generated content and 69,312 comments.\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003e3.3 BERTopic Thematic Model Analysis\u003c/h2\u003e\n\u003cp\u003eThis study uses the BERTopic model for theme identification via text clustering and then conducts sentiment analysis on Weibo posts within each identified theme.\u0026nbsp;This method supports the analysis of rhetorical strategies in CSR advertising content.\u003c/p\u003e\n\u003cp\u003eThe BERTopic model is a natural language processing technique based on BERT (Bidirectional Encoder Representations from Transformers) that automatically constructs topic models from large volumes of text data.\u0026nbsp;In BERTopic, textual content is generally viewed as consisting of related topics and associated words.\u0026nbsp;While a text may encompass multiple topics, each word is assigned to a specific topic.\u0026nbsp;Compared to traditional topic mining methods, BERTopic provides greater insight into latent semantics, particularly excelling in processing short texts, outperforming methods such as Latent Dirichlet Allocation (LDA) and Probabilistic Latent Semantic Analysis (PLSA).\u003c/p\u003e\n\u003cp\u003eThis study begins by using Sentence-BERT for pre-training, generating text embedding vectors for each sentence with a pre-trained BERT model.\u0026nbsp;This process produces a similarity score between -1 and 1, where 1 indicates complete similarity and -1 indicates complete dissimilarity.\u0026nbsp;By vectorizing the text in this way, the vectors of two sentences can be concatenated to measure their similarity.\u0026nbsp;Sentence-BERT captures the underlying semantic relationships between sentences using this method(Reimers \u0026amp; Gurevych, 2019).\u003c/p\u003e\n\u003cp\u003eUMAP (Uniform Manifold Approximation and Projection) is a highly efficient nonlinear dimensionality reduction technique, primarily designed for handling large-scale datasets. The algorithm constructs a local neighborhood graph of the original high-dimensional data, seeking an optimal topological structure in a lower-dimensional space to preserve both global and local structures effectively. The algorithm constructs a local neighborhood graph of the original high-dimensional data, seeking an optimal topological structure in a lower-dimensional space to preserve both global and local structures effectively. In practice, UMAP is widely applied in data visualization, clustering analysis, feature extraction, and anomaly detection. Its flexibility and scalability make it a versatile tool for dimensionality reduction in modern data science and machine learning. (McInnes et al., 2018).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eHDBSCAN (Hierarchical Density-Based Spatial Clustering of Applications with Noise) is an advanced clustering algorithm that identifies data clusters with varying densities and shapes.\u0026nbsp;It performs hierarchical clustering by constructing a minimum spanning tree (MST), which allows it to automatically determine the number of clusters and manage noise effectively.\u0026nbsp;Its robustness enables it to accurately detect primary clustering patterns, even in the presence of outliers and noise.(McInnes \u0026amp; Healy, 2017).\u003c/p\u003e\n\u003cp\u003eFollowing the application of UMAP and HDBSCAN, we employed the Class-based TF-IDF method to extract keywords.\u0026nbsp;Class-based TF-IDF is a variant of the traditional term frequency-inverse document frequency (TF-IDF) method, specifically designed to improve text classification effectiveness by incorporating class information into the calculation.\u0026nbsp;A higher Class-based TF-IDF value indicates a more representative term, making it suitable as a keyword for clustering themes.\u0026nbsp;Based on several experiments and comparisons, our study determined that the optimal number of research topics is seven. (Egger \u0026amp; Yu, 2022).\u003c/p\u003e\n\u003cp\u003eFinally, Top2Vec was employed to merge the most similar topic vectors, reducing the number of topics.Top2Vec combines semantically related topics, effectively reducing the number of topics while preserving semantic relevance.\u0026nbsp;This approach ensures a suitable range of topics by merging them without compromising semantic integrity.\u0026nbsp;Top2Vec accomplishes this by calculating the similarity between topic vectors, identifying the most similar pairs, and merging them into new topic vectors.\u0026nbsp;This process provides optimal topic categorization for the overall BERTopic analysis (Egger \u0026amp; Yu, 2022).\u003c/p\u003e\n\u003cp\u003eThis study focuses on the types of rhetorical strategies used in CSR advertising content to explore their relationship with consumer emotions. However, rhetoric is rooted in linguistic systems structured by human activities(Taboada \u0026amp; Mann,2006). Therefore, this study employs machine learning techniques to cluster datasets by different data collection phases, identifying the most cohesive thematic sentences based on topic dispersion. Subsequently, three linguistics experts were recruited to annotate the samples based on the definitions of rhetorical figures provided in the \u003cem\u003eModern Chinese Dictionary\u003c/em\u003e(Li et al., 2024). Due to the overlap among the 16 rhetorical figures currently classified in Chinese rhetoric, metaphor, simile, and metonymy were grouped into a single category called metaphorical rhetoric to facilitate clustering analysis. The experts independently coded the samples, and in cases of disagreement, they discussed to reach a consensus on each text\u0026apos;s classification (Huang et al., 2022). After annotation, four main clustering themes emerged: three thematic models using rhetorical strategies\u0026mdash;pun, rhetorical question, and metaphor\u0026mdash;and one non-rhetorical model, the statement.\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003e3.4 Sentiment Analysis\u003c/h2\u003e\n\u003cp\u003eSnowNLP is a tool specifically designed for processing Chinese text, implemented on the Python platform, making it particularly suitable for handling Simplified Chinese.\u0026nbsp;It supports various natural language processing functions, including sentiment analysis, word segmentation, part-of-speech tagging, keyword extraction, text classification, and converting text to pinyin.\u0026nbsp;For sentiment analysis, the Chinese text is first passed to the SnowNLP module. SnowNLP examines the words, word relationships, and sentence structures to check for matches in its sentiment dictionary and identify any negative expressions. Once a match is found, the sentiment polarity of the words (positive, neutral, or negative) is determined, and the counts of each type of sentiment word are calculated. Based on the sentiment word counts, their polarity, and the contextual semantics, SnowNLP computes a sentiment score to indicate the emotional intensity of the text. According to the scoring criteria, a score below 0.3 indicates negative sentiment, between 0.3 and 0.6 indicates neutral sentiment, and above 0.6 indicates positive sentiment. SnowNLP then outputs the sentiment score of the text based on this analysis (Chen et al., 2018).\u003c/p\u003e\n\u003ch2\u003e3.5 Theme-Sentiment Evolution Analysis\u003c/h2\u003e\n\u003cp\u003eThis study performs an evolutionary analysis of the rhetorical strategies in CSR advertising text content across two dimensions: theme clustering and sentiment relationships. A theme-sentiment evolution analysis is then conducted based on this framework. By comparing the distribution of themes over different data collection periods, we can map the temporal evolution trajectories of these themes and examine how they relate to sentiment evolution. A Sankey diagram is employed to visualize the results.\u0026nbsp;\u003c/p\u003e"},{"header":"4 Results Analysis","content":"\u003ch2\u003e4.1 Theme Analysis\u003c/h2\u003e\n\u003ch3\u003e4.1.1 Theme analysis before the COVID-19 pandemic.\u003c/h3\u003e\n\u003cp\u003eBefore the COVID-19 pandemic outbreak, 262 Weibo posts and 34,525 comments were collected using Python 3.10. The BERTopic model was employed to extract themes during this period. To enhance interpretability, theme description accuracy, and refine topic modeling results, language experts were invited to summarize and categorize the themes. After summarization, 21 clustered thematic sentences emerged in the pre-pandemic stage, categorized into four types: rhetorical questions, metaphorical rhetoric, puns, and statements (Table1). Statements constituted the majority, with 11 themes accounting for 52.5% of the total, indicating that CSR advertisements predominantly used declarative expressions for consumer communication before the pandemic. Puns and rhetorical questions each had four themes, representing 19% of the total, suggesting that a minority of CSR advertisements used these rhetorical strategies to engage consumers. Finally, metaphorical rhetoric had only two themes, making up 9.5%, indicating that businesses were less inclined to use metaphors for communicating with consumers in CSR advertisements.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAnalysis of thematic keywords using the BERTopic model (Figure 2) reveals that a considerable portion of the content revolves around temporal contexts like \u0026quot;the present,\u0026quot; \u0026quot;now,\u0026quot; and \u0026quot;today.\u0026quot;\u0026nbsp;This indicates that, prior to the COVID-19 pandemic, CSR advertisements mainly focused on the present aspects of people\u0026apos;s lives, paying little attention to future thoughts and plans.\u003c/p\u003e\n\u003cp\u003eTable 1: Themes\u0026nbsp;Before COVID-19\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"552\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"3\" style=\"width: 552px;\"\u003e\n \u003cp\u003eBefore COVID-19\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003eTheme Types\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 313px;\"\u003e\n \u003cp\u003eExamples of Thematic Sentences\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 118px;\"\u003e\n \u003cp\u003eKeywords\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003eDeclarative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 313px;\"\u003e\n \u003cp\u003eKeystone species typically refer to those that play a crucial role in maintaining biodiversity, as well as the structure, function, and stability of ecosystems or biological communities.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 118px;\"\u003e\n \u003cp\u003eDiversity/Ecosystem /Biological/Function\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003eQuestion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 313px;\"\u003e\n \u003cp\u003e\u0026nbsp;What can we do to love the Earth? Today, share for the planet: choose low-carbon transportation, save water, reduce plastic use, clean your plate, and sort waste. Let\u0026apos;s take action together!\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 118px;\"\u003e\n \u003cp\u003eToday/Earth/Action\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003ePun\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 313px;\"\u003e\n \u003cp\u003eProtect the environment, let\u0026apos;s take \u0026apos;art\u0026apos;-ion together!\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 118px;\"\u003e\n \u003cp\u003eProtect/Action/Environment\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003eDeclarative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 313px;\"\u003e\n \u003cp\u003eDon\u0026rsquo;t let the beauty of Earth exist only in images; don\u0026rsquo;t let nature and vitality remain only in memories. Humanity should not only explore the future but also join hands to protect the present.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 118px;\"\u003e\n \u003cp\u003eNow/Protect/Nature\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003eDeclarative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 313px;\"\u003e\n \u003cp\u003eLet us focus on the importance of climate change and biodiversity conservation, and raise our sense of responsibility. Let\u0026apos;s be friends of nature together.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 118px;\"\u003e\n \u003cp\u003eClimate Change/Biological/Conservation\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003eDeclarative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 313px;\"\u003e\n \u003cp\u003eLet us respect nature, protect nature, and work together to build a community of life between humans and nature to address the challenges of climate change.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 118px;\"\u003e\n \u003cp\u003eClimate Change/Respect/Nature\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003eDeclarative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 313px;\"\u003e\n \u003cp\u003eProtecting nature and caring for the environment are not just slogans for today but actions for the future.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 118px;\"\u003e\n \u003cp\u003eToday/Protect/Eenvironment\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003eMetaphor\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 313px;\"\u003e\n \u003cp\u003eLet the Earth thrive with an extra layer, like a wooden coat, for healthy growth\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 118px;\"\u003e\n \u003cp\u003eEarth/Health\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003eQuestion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 313px;\"\u003e\n \u003cp\u003eIn your daily life, what small things do you consistently do for environmental protection?\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 118px;\"\u003e\n \u003cp\u003eDaily Life/Protect\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003eQuestion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 313px;\"\u003e\n \u003cp\u003eDo you wish every drop of water could be clean? Then start by saving water.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 118px;\"\u003e\n \u003cp\u003eClean/Save\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003eDeclarative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 313px;\"\u003e\n \u003cp\u003eCommitted to reducing carbon emissions and promoting green development, our innovative technologies not only change lives but also protect the Earth. Let\u0026apos;s work together for a sustainable future!\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 118px;\"\u003e\n \u003cp\u003eCommitted to/Innovation\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003eDeclarative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 313px;\"\u003e\n \u003cp\u003eRemember the source when you drink water, and respect nature! Baisui Mountain is always committed to protecting our water sources.\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 118px;\"\u003e\n \u003cp\u003eProtect/Committed to\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003eDeclarative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 313px;\"\u003e\n \u003cp\u003eLet us join hands now to protect nature and work together to slow down the trend of global warming.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 118px;\"\u003e\n \u003cp\u003eNow/Protect/Nature\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003eQuestion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 313px;\"\u003e\n \u003cp\u003eDid you know? Approximately 7 million people die each year from air pollution, at least 60,000 species go extinct annually, and 6 million hectares of land turn into deserts every year... Let\u0026rsquo;s take action together to protect our clear waters and green mountains!\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 118px;\"\u003e\n \u003cp\u003ePollution/Action/Now\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003eDeclarative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 313px;\"\u003e\n \u003cp\u003eLet\u0026apos;s join hands to protect the ecological environment and live a healthy life together.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 118px;\"\u003e\n \u003cp\u003eEnvironment/Join Hands/Co-create\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003eDeclarative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 313px;\"\u003e\n \u003cp\u003eAt this very moment, let\u0026apos;s join the world in turning off the lights for one hour, taking action towards a more environmentally friendly lifestyle.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 118px;\"\u003e\n \u003cp\u003eRight now/Action/Committed to\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003eDeclarative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 313px;\"\u003e\n \u003cp\u003eProtecting wildlife reflects humane care and values.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 118px;\"\u003e\n \u003cp\u003eWildlife/Humane Care\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003ePun\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 313px;\"\u003e\n \u003cp\u003eHarnessing \u0026quot;nuclear\u0026quot; power to protect our green waters and mountains, supporting nature\u0026apos;s wellbeing, we are taking action!\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 118px;\"\u003e\n \u003cp\u003eNature/Action\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003ePun\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 313px;\"\u003e\n \u003cp\u003eUse eco-friendly travel products on your journey. Be a responsible traveler and protect the ecosystem through small actions!\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 118px;\"\u003e\n \u003cp\u003eEnvironment/Responsibility/Protection/Ecosystem/Action\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003ePun\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 313px;\"\u003e\n \u003cp\u003eReduce a \u0026quot;carbon\u0026quot; footprint today, and receive nature\u0026apos;s rewards tomorrow.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 118px;\"\u003e\n \u003cp\u003eEarth/ Nature/ Today\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 121px;\"\u003e\n \u003cp\u003eMetaphor\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 313px;\"\u003e\n \u003cp\u003eTogether, let\u0026apos;s love the Earth and live in harmony with nature. We are all children of the natural world.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 118px;\"\u003e\n \u003cp\u003eEarth/ Nature/Love Together\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003ch3\u003e4.1.2 Theme analysis after the COVID-19 pandemic\u003c/h3\u003e\n\u003cp\u003eDuring the post-COVID-19 pandemic period, 246 blog posts and 34,787 comments were collected using Python 3.10.\u0026nbsp;Following expert summarization and categorization, 25 clustered thematic sentences were identified, categorized into four types: rhetorical questions, metaphorical rhetoric, puns, and declarative expressions (Table 2).\u0026nbsp;Declarative expressions accounted for 8 of the 25 themes, marking a significant decrease from 52.5% to 32%.\u0026nbsp;Rhetorical questions made up 3 themes, reducing their share from 19% to 12%.\u0026nbsp;Puns emerged as the dominant theme, accounting for 10 sentences and increasing from 19% to 40%.\u0026nbsp;Finally, metaphorical rhetoric rose from 2 to 4 themes, raising its proportion to 16%.\u003c/p\u003e\n\u003cp\u003eBy analyzing temporal evolution, we identified new trends when comparing pre- and post-pandemic samples. A growing number of CSR advertisements are adopting various rhetorical theme types to engage with consumers. Notably, after the COVID-19 pandemic, the use of puns as a theme has significantly increased, becoming the preferred rhetorical strategy in many CSR advertising texts today.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThrough post-pandemic BERTopic theme clustering analysis, we discovered that CSR advertisements have shifted from focusing on \u0026quot;present\u0026quot; lifestyle discussions to emphasizing \u0026quot;future\u0026quot; considerations and actions. Additionally, more CSR advertisements are employing rhetorical strategies to refine their content, with the majority choosing puns as their primary rhetorical device. This trend underscores the growing attention and utilization of puns by CSR advertisements in recent years.\u003c/p\u003e\n\u003cp\u003eTable 2: Themes After COVID-19\u003c/p\u003e\n\u003cdiv align=\"\"\u003e\n \u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"557\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"3\" style=\"width: 557px;\"\u003e\n \u003cp\u003eAfter COVID-19\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003eTheme Types\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003eExamples of Thematic Sentences\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eKeywords\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003ePun\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003e\u0026quot;GOOD GOOD, Let\u0026apos;s make the Earth less \u0026apos;carbonated\u0026apos; and take action towards the future.\u0026quot;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eCarbon Gas/Action/Future\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003ePun\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003eEarth Day AI sci-fi blockbuster, Jane AI Earth in the distant future, what will happen to the Earth depends on our choices today, Jane AI Earth, man and nature live in harmony.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eAI/Cherish/Future/Earth/Nature\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003ePun\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003eFor the sake of Earth\u0026apos;s tomorrow, I\u0026rsquo;ll stop \u0026ldquo;straw\u0026rdquo;ing! Starting now, let\u0026rsquo;s stop using straws for the Earth!\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eStraws/Earth/Tomorrow\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003ePun\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003eGreen living, shared future, let the Earth \u0026ldquo;chick\u0026rdquo;en up with vitality.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eFuture/Earth/Life\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003ePun\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003eLess carbon \u0026quot;sighs,\u0026quot; more TREES; join us in adding a touch of \u0026quot;green magic\u0026quot; to the Earth.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eCarbon footprint/Green initiatives/Earth\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003eMetaphor\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003eWe are nature\u0026rsquo;s children, protecting nature is protecting our future selves.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eNature/Children/Tomorrow/Protection\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003eMetaphor\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003eEarth, the mother of humanity, the cradle of life, is so beautiful and majestic, yet so kind and gentle. It is only this big and won\u0026apos;t grow any bigger. If it is destroyed, we have nowhere else to go.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eEarth/Mother/Destruction/Cradle/Life\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003ePun\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003eDaily dark chocolate, smartly protect the Earth, the future is shaped by our actions.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eSmart Protection/Earth/Future\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003eMetaphor\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003eMaster the future, embrace electric power. Let\u0026rsquo;s charge the Earth together.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eFuture/Earth/Recharge\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003eQuestion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003eWithout a low-carbon lifestyle, can our Earth still be healthy?\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eLife/Low-carbon/Health/Earth\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003eMetaphor\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003eEvery small \u0026quot;cold\u0026quot; of Mother Earth is a major disaster for all of humanity.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eDisaster/Humanity/Earth/Mother\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003eDeclarative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003eCherish the Earth, we are always in action.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eCherish/Earth/Action\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003ePun\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003eEating McDonald\u0026rsquo;s, don\u0026rsquo;t forget the small efforts we can make; let\u0026rsquo;s be environmentally conscious together towards the future.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eSmall Efforts/Environmental Protection\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003eDeclarative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003eBe a friend of nature; take an hour to do something beneficial for the Earth. Let\u0026apos;s protect a better future for nature, hand in hand.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eNature/Friends/Hand in Hand\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003ePun\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003e\u0026quot;Guarding\u0026quot; the Earth: Buying groceries? Why so much plastic? One eco-friendly bag is enough!\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eNo Plastic/Environmental/Guardianship/Earth\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003eDeclarative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003eGuardians of the national park use their hands to protect this beautiful yet fragile natural environment.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eNational Park/Nature/Environment/Protection\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003eQuestion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003eIn the warm spring, around the time of awakening insects. Do you know how important the finless porpoise is to the health of the Yangtze River ecosystem? As the only remaining freshwater porpoise in the Yangtze, a first-class national protected wild animal, and a unique rare freshwater porpoise in China, its population is slowly recovering, but these creatures, with their healing smiles, still face severe survival threats.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eEcosystem/Health/Wildlife\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003eDeclarative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003eLoving the Earth is not just about Earth Day; let\u0026rsquo;s adopt daily plastic reduction practices for a future-oriented lifestyle\u0026mdash;reduce the use of plastic products.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eEart/ Future/Plastic Reduction\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003eDeclarative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003eSadly, since the pandemic, our Earth has faced a new form of pollution: mask pollution. Billions of discarded masks have been thrown away, entering the natural environment and becoming a nightmare for wildlife. We hope everyone will store masks properly and discard them at medical waste collection points in hospitals. Protecting the Earth is not just about slogans on Earth Day, but starts with small everyday actions.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003ePandemic/Pollution/Masks/Nightmare\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003eQuestion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003eCan you find where the nature that represents positive change is? Let\u0026rsquo;s continue to protect biodiversity together and do something beneficial for the home we all depend on. Let\u0026rsquo;s gather, support each other, and be grateful for the beautiful home we have.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eNature/Protection/Biodiversity/Survival\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003eDeclarative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003eLet\u0026rsquo;s join forces to keep life shining brightly in nature. With gentle strength, let\u0026rsquo;s preserve the beauty of nature so that we have no regrets tomorrow.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eTomorrow/Life/Strength/Radiant/Nature\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003eDeclarative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003eHappy Earth Day to our only home.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eHome/Earth/Joy\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003eDeclarative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003eLandmarks around the world will turn off their lights; facing the future, we are stronger together.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eLights Out/Future/Strength\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003ePun\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003eThe Low-Carbon Diary of the Great Sage Wins the Cycling Season. Choose green and low-carbon travel, protect the blue sky and white clouds, feel the gifts of nature, and make a promise to the future.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eGreen Warrior/Low-carbo/ Future/Nature\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 134px;\"\u003e\n \u003cp\u003ePun\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 291px;\"\u003e\n \u003cp\u003ePlastic reduction zone, please slow down on plastic use. Let\u0026rsquo;s speak up for environmental protection together.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eReduce Plastic/Environmental Protection\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003ch2\u003e4.2 Sentiment Analysis\u003c/h2\u003e\n\u003cp\u003eFollowing the completion of topic modeling, this study utilized SnowNLP to perform sentiment analysis on thematic sentences for each theme type, based on temporal evolution (Figure 2).\u0026nbsp;As per the sentiment scoring settings of SnowNLP, the pun theme scored 0.71 pre-pandemic, rising to 0.75 post-pandemic, an increase of 0.04.\u0026nbsp;This suggests that the pun theme has gained increasing popularity among consumers in recent years.\u0026nbsp;The declarative expression theme scored 0.70 pre-pandemic but dropped significantly to 0.60 post-pandemic, indicating a substantial decline in sentiment.\u0026nbsp;This implies that, following the onset of COVID-19, consumers have become weary and even resentful of CSR advertisement content lacking rhetorical strategies.\u0026nbsp;Both metaphorical rhetoric and rhetorical question themes maintained neutral sentiment, with scores slightly increasing from 0.50 and 0.44 pre-pandemic to 0.52 and 0.50 post-pandemic, respectively.\u0026nbsp;The minor increases in sentiment scores indicate that the pandemic did not significantly heighten consumer interest in these rhetorical themes.\u003c/p\u003e\n\u003ch2\u003e4.3 Theme-Sentiment Evolution Analysis\u003c/h2\u003e\n\u003cp\u003eThis study examines the rhetorical strategies in CSR advertising texts from both thematic and emotional perspectives. Based on this analysis, a Sankey diagram was created to illustrate the evolution of themes and sentiments, as depicted in Figure 3. The diagram illustrates the emotional flow of CSR advertising rhetorical themes across two distinct periods: pre-COVID-19 and post-COVID-19. Additionally, Tables 3 and 4 show the sentiment distribution proportions of themes for each period.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ePrior to the COVID-19 pandemic, as indicated in Table 3, declarative themes were the dominant mode of expression in CSR advertising texts. These themes exhibited a relatively high proportion of positive sentiment, suggesting that consumers were receptive to companies using this method to convey CSR messages. The proportions of neutral and negative sentiment further imply that consumers did not oppose this mode of expression. Previous research noted that pre-pandemic CSR advertising strategies frequently employed language emphasizing \u0026quot;now\u0026quot; and \u0026quot;the present,\u0026quot; reflecting a shared effort by consumers and CSR advertisements to create an immediate consumption discourse environment. Declarative expressions, which convey information straightforwardly, effectively aligned with this immediate consumption context. Consequently, declarative expressions not only made up more than half of the thematic quantity compared to other themes during this period, but the high proportion of positive consumer sentiment also clearly demonstrated an interest in and preference for this mode of CSR advertising.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAs shown in Figure 4, the proportion of the use of puns and rhetorical questions as rhetorical themes during this period is equal. However, the emotional distribution among consumers differs significantly. Puns evoke the most positive emotions among consumers, whereas rhetorical questions evoke the highest proportion of negative emotions. These two rhetorical themes, with the same frequency but opposite consumer reactions, indicate that consumers prefer a more humorous and witty approach in green brand communication, which better aligns with their psychological preferences. Rhetorical questions primarily use techniques such as questioning to generate text content, often perceived as didactic and containing elements of preaching, such as \u0026quot;Do you want every drop of water to be clean? Then start saving water.\u0026quot; Consumers tend to feel aversion and dissatisfaction towards such content.\u003c/p\u003e\n\u003cp\u003eMetaphorical rhetoric themes were relatively underrepresented during the pre-COVID-19 period. Consumers predominantly expressed neutral sentiment towards these themes, which also had the lowest proportion of negative sentiment among the four theme types. Although metaphorical rhetoric was not frequently used as the primary textual strategy in CSR advertisements, its status as the theme with the least negative sentiment suggests it remains a preferred choice for CSR campaigns seeking to gauge consumer reactions.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAs shown in Table 4, following the COVID-19 pandemic, the proportion of declarative themes dropped significantly from 52.5% to 32%, marking a decrease of 39%. The proportion of positive sentiment similarly decreased from 0.6402 to 0.4195, dropping from the second to the third position. Meanwhile, negative sentiment rose from the third to the second position, whereas neutral sentiment remained relatively stable. These shifts in theme and sentiment proportions suggest that both CSR advertisers and consumers have become less receptive to straightforward declarative content after the pandemic. From the aggregated thematic keywords discussed earlier, it is evident that \u0026quot;future\u0026quot; and \u0026quot;tomorrow\u0026quot; have emerged as central contexts for themes post-pandemic, indicating a shift from immediate consumption to discussions about future lifestyles. This shift in contextual focus in CSR advertising has contributed to the decline of declarative expressions, as both advertisers and consumers now favor content that addresses the long-term impact of their actions. Consequently, consumers have grown increasingly indifferent to declarative themes.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAs illustrated in Figure 5, pun themes have emerged as the preferred choice for most CSR advertisements on social media, with their proportion increasing from 19% pre-pandemic to 40% post-pandemic. The positive sentiment proportion for pun themes consistently remains the highest among all theme types, maintaining a score above 0.7. This indicates that CSR advertisements are increasingly adopting puns for consumer engagement, with consumers appreciating this humorous and highly engaging rhetorical style. Examples include statements like, \u0026quot;For the sake of the Earth\u0026apos;s tomorrow, I don\u0026apos;t \u0026apos;straw\u0026apos; it! From now on, for the Earth, no more straws!\u0026quot; and \u0026quot;GOOD GOOD, reduce \u0026apos;carbon\u0026apos; emissions, act for the future.\u0026quot;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn the post-pandemic period, the proportion of metaphorical rhetoric themes increased from 9.5% to 16%. Although sentiment proportions did not change significantly compared to the pre-pandemic period, metaphorical rhetoric themes continued to show the lowest negative sentiment and the highest neutral sentiment among all themes. This indicates that some CSR advertisements are exploring consumer emotional attitudes and adjusting their emotional connections with consumers.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe proportion of rhetorical question themes has decreased from 19% to 12%, making it the least used theme type in CSR advertisements. Not only have CSR advertisements gradually abandoned this theme, but consumers continue to show aversion towards it. The proportion of negative sentiment has not improved despite the shift towards \u0026quot;future\u0026quot; oriented content, making it still the most disliked rhetorical theme among consumers.\u003c/p\u003e\n\u003cp\u003eThe distribution of rhetorical types after the COVID-19 pandemic has shown significant changes compared to before the pandemic, as depicted in Figures 4 and 5. In positive emotions, the proportion of declarative has decreased, while the use of pun and metaphor has increased. This shift may reflect a tendency during the pandemic for people to employ humorous and symbolic expressions to alleviate stress. Conversely, in negative emotions, there has been a notable increase in the proportion of questions, possibly related to the uncertainties, challenges, and difficulties brought about by the pandemic, as people express doubts and concerns through questioning. Furthermore, examining rhetorical changes across emotional categories reveals varied impacts of the COVID-19 on rhetorical proportions. In positive emotions, there appears to be greater diversity in rhetorical types, with an increase not only in direct declarative expressions but also in the use of pun and metaphor to enrich emotional expression. In neutral emotions, changes in rhetorical types have been relatively stable, yet there has been an increase in the use of metaphor, possibly reflecting a heightened preference for symbolic and metaphorical expressions during the pandemic. In negative emotions, the increase in the proportion of questions is particularly pronounced, further emphasizing the negative impact of the pandemic on people\u0026apos;s lives and mental health.\u003c/p\u003e\n\u003cp\u003eIn summary, the COVID-19 pandemic has significantly influenced the distribution of rhetorical types. The increase in pun and metaphor in positive emotions reflects a trend towards seeking humor and symbolic expression during the pandemic. The notable increase in questions in negative emotions highlights the pandemic\u0026apos;s adverse effects on people\u0026apos;s lives and mental health. These changes not only reflect the pandemic\u0026apos;s impact on rhetorical usage but also reveal shifts in people\u0026apos;s psychological responses and adaptation processes during this challenging time.\u003c/p\u003e\n\u003cp\u003eTable 3 Theme Sentiment Distribution Before the COVID-19 Pandemic\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"100%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eTheme Type\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eProportion of Theme Quantity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 20px;\"\u003e\n \u003cp\u003eProportion of Positive Sentiment\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 22px;\"\u003e\n \u003cp\u003eProportion of Neutral Sentiment\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 23px;\"\u003e\n \u003cp\u003eProportion of Negative Sentiment\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eDeclarative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003e52.50%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 20px;\"\u003e\n \u003cp\u003e0.6402\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 22px;\"\u003e\n \u003cp\u003e0.2682\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 23px;\"\u003e\n \u003cp\u003e0.0916\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003ePun\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003e19%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 20px;\"\u003e\n \u003cp\u003e0.7133\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 22px;\"\u003e\n \u003cp\u003e0.1515\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 23px;\"\u003e\n \u003cp\u003e0.135\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eQuestion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003e19%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 20px;\"\u003e\n \u003cp\u003e0.175\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 22px;\"\u003e\n \u003cp\u003e0.5225\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 23px;\"\u003e\n \u003cp\u003e0.3025\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eMetaphor\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003e9.50%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 20px;\"\u003e\n \u003cp\u003e0.124\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 22px;\"\u003e\n \u003cp\u003e0.7895\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 23px;\"\u003e\n \u003cp\u003e0.0865\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eTable 4 Theme Sentiment Distribution After the COVID-19 Pandemic\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"100%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eTheme Type\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eProportion of Theme Quantity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 20px;\"\u003e\n \u003cp\u003eProportion of Positive Sentiment\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 22px;\"\u003e\n \u003cp\u003eProportion of Neutral Sentiment\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 23px;\"\u003e\n \u003cp\u003eProportion of Negative Sentiment\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eDeclarative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 17px;\"\u003e\n \u003cp\u003e32%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e0.4195\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e0.4169\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 23px;\"\u003e\n \u003cp\u003e0.1636\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003ePun\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 17px;\"\u003e\n \u003cp\u003e40%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e0.7082\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e0.2237\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 23px;\"\u003e\n \u003cp\u003e0.0681\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eQuestion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 17px;\"\u003e\n \u003cp\u003e12%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e0.3037\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e0.4537\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 23px;\"\u003e\n \u003cp\u003e0.2427\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eMetaphor\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 17px;\"\u003e\n \u003cp\u003e16%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e0.1788\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e0.7725\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 23px;\"\u003e\n \u003cp\u003e0.0488\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e"},{"header":"5 Discussion of Research Findings","content":"\u003col class=\"decimal_type\"\u003e\n \u003cli\u003eIn terms of text theme mining, this paper analyzes the microblog texts of the pre-New Crown epidemic and post-New Crown epidemic phases, and identifies the shift of CSR advertisements from declarative to punning rhetorical themes in generating textual content, and according to the aggregated data on theme word extraction, it can be found that the expression of textual content has changed from the present to the future, and this change also proves that the major public health emergency has made CSRs focus more on \u0026ldquo;consumers\u0026apos; present behavior will affect the future of society\u0026rdquo; when constructing advertisements. The change from \u0026ldquo;present\u0026rdquo; to \u0026ldquo;future\u0026rdquo; in creating the context of text content is also evidence that the emergence of a major public health event has made CSRs focus more on \u0026ldquo;how consumer behavior in the present will affect society in the future\u0026rdquo; in constructing the content of advertisements.\u003c/li\u003e\n \u003cli\u003eIn terms of sentiment analysis, consumers\u0026apos; sentiment scores for CSR advertisements are in the distribution range of neutral sentiment share, and from the scores of the pre-stage and post-stage of the New Crown Epidemic, the pre-stage is 0.59, and the post-stage score is 0.60, which can indicate that after the New Crown Epidemic, consumers pay more attention to and like the CSR advertisements and it also indicates that the suddenness of the major public health event from the side makes consumers have positive emotions towards CSR advertisements and the sustainable behaviors they advocate.\u003c/li\u003e\n \u003cli\u003eIn terms of topic-emotion evolution analysis, before the Xin Guan epidemic, most CSR advertisements preferred to use statement expressions, which are simple, clear in purpose, and do not contain rhetorical content, for marketing communication on social platforms, and consumers have more positive emotional expressions for them, but with the changes in the social environment, the original topic context and type of topic expression gradually changed when CSR advertisements generated text content on social platforms. However, with the change of social environment, CSR advertisements have gradually changed the original topic context and theme expression type when generating text content on social platforms, and the number of CSR advertisements containing declarative themes has declined, and has been replaced by punning rhetorical themes as the main rhetorical mode of CSR advertisements in social media marketing communication. Consumers are more likely to accept CSR advertisements that use humorous semantics and attention-grabbing language styles than statements. Therefore, punning rhetoric has become the main way of marketing communication of CSR ads to consumers in the post epidemic era.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"6 Research Significance and Shortcomings","content":"\u003cp\u003eThis study utilized the BERTopic model to conduct theme clustering analysis of the rhetorical content in CSR advertisements on Chinese social media platforms and employed SnowNLP for sentiment scoring.\u0026nbsp;Based on previous research findings, this study first demonstrated that the rhetorical themes in CSR advertisement content shifted before and after the COVID-19 pandemic. Puns have become the commonly used rhetorical strategy in CSR ads, and consumers maintain a positive emotional attitude towards this theme type. This finding fills the gap in CSR research concerning the relationship between CSR advertising and consumer emotions in the context of Chinese language, providing a basis for further study.\u0026nbsp;Secondly, due to the COVID-19 pandemic, the semantic context on Chinese social media shifted from focusing on the \u0026quot;present\u0026quot; to considering the \u0026quot;future.\u0026quot; This shift led to a significant increase in the use of rhetorical strategies in CSR advertisements, demonstrating that unexpected events can significantly impact the relationship between online semantic contexts and CSR advertising communication.\u003c/p\u003e\n\u003cp\u003eThis in-depth study of CSR advertising at different stages has established a relationship between CSR advertisements and changes in consumer emotions, enriching related research. However, several limitations remain: 1. This study only compared data samples from 2017-2019 and 2021-2023, without conducting long-term tracking. Future research should consider changes over a longer time scale. 2. This study used only the BERTopic clustering model and SnowNLP sentiment analysis, both suitable for short text processing. Since CSR advertising involves more than just Weibo posts, future studies could incorporate other methods for comparative analysis.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eJ.W. conceptualized the study, conducted the data analysis, and drafted the initial version of the manuscript. X.S. participated in the study design, assisted with data interpretation, and critically revised the manuscript. W.Z., as the corresponding author, contributed to the writing and editing of the manuscript. Y.J. was responsible for data cleansing and collation, while Y.L. verified and proofread the data. All authors reviewed and approved the final manuscript.Informed consent were obtained from all the paticipants.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003eThe authors would like to express their gratitude to the National Natural Science Foundation of China ( No. 72102172) for their financial support of this research.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eData Availability StatementThe data that support the findings of this study are not publicly available as they are part of an ongoing research project. For any inquiries regarding the data, please contact the corresponding author（Dr. wei ZHANG / [email protected]）. The data will be provided if necessary for review purposes or specific requests.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eAngelov, D. (2020). Top2Vec: Distributed Representations of Topics (arXiv:2008.09470). arXiv. http://arxiv.org/abs/2008.09470\u003c/li\u003e\n \u003cli\u003eBergkvist, L., \u0026amp; Zhou, K. Q. (2019). Cause-related marketing persuasion research: An integrated framework and directions for further research. International Journal of Advertising, 38(1), 5\u0026ndash;25. https://doi.org/10.1080/02650487.2018.1452397\u003c/li\u003e\n \u003cli\u003eCai, M. (2021). Natural language processing for urban research: A systematic review. Heliyon, 7(3), e06322. https://doi.org/10.1016/j.heliyon.2021.e06322\u003c/li\u003e\n \u003cli\u003eCarroll, A. B. (2021). Corporate social responsibility (CSR) and the COVID-19 pandemic: Organizational and managerial implications. Journal of Strategy and Management, 14(3), 315\u0026ndash;330. https://doi.org/10.1108/JSMA-07-2021-0145\u003c/li\u003e\n \u003cli\u003eChang, C. (2012). The effectiveness of advertising that leverages sponsorship and cause-related marketing: A contingency model. International Journal of Advertising, 31(2), 317\u0026ndash;338. https://doi.org/10.2501/IJA-31-2-317-337\u003c/li\u003e\n \u003cli\u003eChen, C., Chen, J., \u0026amp; Shi, C. (2018). Research on Credit Evaluation Model of Online Store Based on SnowNLP. E3S Web of Conferences, 53, 03039. https://doi.org/10.1051/e3sconf/20185303039\u003c/li\u003e\n \u003cli\u003eChernev, A., \u0026amp; Blair, S. (2015). Doing Well by Doing Good: The Benevolent Halo of Corporate Social Responsibility. Journal of Consumer Research, 41(6), 1412\u0026ndash;1425. https://doi.org/10.1086/680089\u003c/li\u003e\n \u003cli\u003eChoi, C. S., Cho, Y.-N., Ko, E., Kim, S. J., Kim, K. H., \u0026amp; Sarkees, M. E. (2019). Corporate sustainability efforts and e-WOM intentions in social platforms. International Journal of Advertising, 38(8), 1224\u0026ndash;1239. https://doi.org/10.1080/02650487.2019.1613858\u003c/li\u003e\n \u003cli\u003eChoi, C.-W. (2022). Increasing company-cause fit: The effects of a relational ad message and consumers\u0026rsquo; cause involvement on attitude toward the CSR activity. International Journal of Advertising, 41(2), 333\u0026ndash;353. https://doi.org/10.1080/02650487.2020.1822060\u003c/li\u003e\n \u003cli\u003eDiehl, S., Terlutter, R., \u0026amp; Mueller, B. (2016). Doing good matters to consumers: The effectiveness of humane-oriented CSR appeals in cross-cultural standardized advertising campaigns. International Journal of Advertising, 35(4), 730\u0026ndash;757. https://doi.org/10.1080/02650487.2015.1077606\u003c/li\u003e\n \u003cli\u003eDitlev-Simonsen, C., \u0026amp; Wenst\u0026oslash;p, S. (n.d.). Companies\u0026rsquo; ethical commitment: An analysis of the rhetoric in CSR reports.\u003c/li\u003e\n \u003cli\u003eEberle, D., Berens, G., \u0026amp; Li, T. (2013). The Impact of Interactive Corporate Social Responsibility Communication on Corporate Reputation. Journal of Business Ethics, 118(4), 731\u0026ndash;746. https://doi.org/10.1007/s10551-013-1957-y\u003c/li\u003e\n \u003cli\u003eEgger, R., \u0026amp; Yu, J. (2022). A Topic Modeling Comparison Between LDA, NMF, Top2Vec, and BERTopic to Demystify Twitter Posts. Frontiers in Sociology, 7, 886498. https://doi.org/10.3389/fsoc.2022.886498\u003c/li\u003e\n \u003cli\u003eFeldner, S. B. (2014). How Corporations Manage Industry and Consumer Expectations via the CSR Report. Public Relations Journal, 8(3).\u003c/li\u003e\n \u003cli\u003eFriestad, M., \u0026amp; Wright, P. (1994). The Persuasion Knowledge Model: How People Cope with Persuasion Attempts. Journal of Consumer Research, 21(1), 1\u0026ndash;31. https://doi.org/10.1086/209380\u003c/li\u003e\n \u003cli\u003eGao, Y. (2011). CSR in an emerging country: A content analysis of CSR reports of listed companies. Baltic Journal of Management, 6(2), 263\u0026ndash;291. https://doi.org/10.1108/17465261111131848\u003c/li\u003e\n \u003cli\u003eGrootendorst, M. (2022). BERTopic: Neural topic modeling with a class-based TF-IDF procedure (arXiv:2203.05794). arXiv. http://arxiv.org/abs/2203.05794\u003c/li\u003e\n \u003cli\u003eHartmann, P., Marcos, A., \u0026amp; Apaolaza, V. (2023). Past, present, and future of research on corporate social responsibility advertising. International Journal of Advertising, 42(1), 87\u0026ndash;95. https://doi.org/10.1080/02650487.2022.2121902\u003c/li\u003e\n \u003cli\u003eHu, N. (2022). Sentiment Analysis of Texts on Public Health Emergencies Based on Social Media Data Mining. Computational and Mathematical Methods in Medicine, 2022, 1\u0026ndash;7. https://doi.org/10.1155/2022/3964473\u003c/li\u003e\n \u003cli\u003eHuang, C., Zhuang, S., Li, Z., \u0026amp; Gao, J. (2022). Creating a Sincere Sustainable Brand: The Application of Aristotle\u0026rsquo;s Rhetorical Theory to Green Brand Storytelling. Frontiers in Psychology, 13, 897281. https://doi.org/10.3389/fpsyg.2022.897281\u003c/li\u003e\n \u003cli\u003eJing Wen, T., Kim, E., Wu, L., \u0026amp; Dodoo, N. A. (2020). Activating persuasion knowledge in native advertising: The influence of cognitive load and disclosure language. International Journal of Advertising, 39(1), 74\u0026ndash;93. https://doi.org/10.1080/02650487.2019.1585649\u003c/li\u003e\n \u003cli\u003eLee, S. Y., \u0026amp; Carroll, C. E. (2011). The Emergence, Variation, and Evolution of Corporate Social Responsibility in the Public Sphere, 1980\u0026ndash;2004: The Exposure of Firms to Public Debate. Journal of Business Ethics, 104(1), 115\u0026ndash;131. https://doi.org/10.1007/s10551-011-0893-y\u003c/li\u003e\n \u003cli\u003eLi, S. (Kevin), Wang, L., Huang, J., Gopal, R., \u0026amp; Lin, Z. (2024). Sizzle or Fizzle? Supply and Consumption Dynamics of Home-Cooked Food on Sharing Platform. Production and Operations Management, 33(1), 11\u0026ndash;31. https://doi.org/10.1177/10591478231224954\u003c/li\u003e\n \u003cli\u003eLiu, C., Chen, S., \u0026amp; Shao, Q. (2019). Do CEO Rhetorical Strategies Affect Corporate Social Performance? Evidence from China. Sustainability, 11(18), 4907. https://doi.org/10.3390/su11184907\u003c/li\u003e\n \u003cli\u003eLloyd-Smith, P., \u0026amp; An, H. (2019). Are corporate social responsibility and advertising complements or substitutes in producing firm reputation? Applied Economics, 51(21), 2275\u0026ndash;2288. https://doi.org/10.1080/00036846.2018.1540858\u003c/li\u003e\n \u003cli\u003eLowrey, T. M. (2002). Does language matter? Psychology \u0026amp; Marketing, 19(7\u0026ndash;8), 569\u0026ndash;572. https://doi.org/10.1002/mar.10025\u003c/li\u003e\n \u003cli\u003eLuangrath, A. W., Peck, J., \u0026amp; Barger, V. A. (2017). Textual paralanguage and its implications for marketing communications. Journal of Consumer Psychology, 27(1), 98\u0026ndash;107. https://doi.org/10.1016/j.jcps.2016.05.002\u003c/li\u003e\n \u003cli\u003eMarais, M. (2012). CEO rhetorical strategies for corporate social responsibility (CSR). Society and Business Review, 7(3), 223\u0026ndash;243. https://doi.org/10.1108/17465681211271314\u003c/li\u003e\n \u003cli\u003eMcInnes, L., \u0026amp; Healy, J. (2017). Accelerated Hierarchical Density Based Clustering. 2017 IEEE International Conference on Data Mining Workshops (ICDMW), 33\u0026ndash;42. https://doi.org/10.1109/ICDMW.2017.12\u003c/li\u003e\n \u003cli\u003eMcInnes, L., Healy, J., Saul, N., \u0026amp; Gro\u0026szlig;berger, L. (2018). UMAP: Uniform Manifold Approximation and Projection. Journal of Open Source Software, 3(29), 861. https://doi.org/10.21105/joss.00861\u003c/li\u003e\n \u003cli\u003ePanagiotopoulos, I. (2021). Novel CSR \u0026amp; novel coronavirus: Corporate social responsibility inside the frame of coronavirus pandemic in Greece. International Journal of Corporate Social Responsibility, 6(1), 10. https://doi.org/10.1186/s40991-021-00065-7\u003c/li\u003e\n \u003cli\u003ePogacar, R., Shrum, L. J., \u0026amp; Lowrey, T. M. (2018). The Effects of Linguistic Devices on Consumer Information Processing and Persuasion: A Language Complexity \u0026times; Processing Mode Framework. Journal of Consumer Psychology, 28(4), 689\u0026ndash;711. https://doi.org/10.1002/jcpy.1052\u003c/li\u003e\n \u003cli\u003eRahman, M., Rodr\u0026iacute;guez-Serrano, M. \u0026Aacute;., \u0026amp; Lambkin, M. (2017). Corporate Social Responsibility And Marketing Performance: The Moderating Role Of Advertising Intensity. Journal of Advertising Research, 57(4), 368\u0026ndash;378. https://doi.org/10.2501/JAR-2017-047\u003c/li\u003e\n \u003cli\u003eReimers, N., \u0026amp; Gurevych, I. (2019). Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks (arXiv:1908.10084). arXiv. http://arxiv.org/abs/1908.10084\u003c/li\u003e\n \u003cli\u003eS\u0026aacute;nchez‐Franco, M. J., \u0026amp; Rey‐Moreno, M. (2022). Do travelers\u0026rsquo; reviews depend on the destination? An analysis in coastal and urban peer‐to‐peer lodgings. Psychology \u0026amp; Marketing, 39(2), 441\u0026ndash;459. https://doi.org/10.1002/mar.21608\u003c/li\u003e\n \u003cli\u003eSchouten, A. P., Janssen, L., \u0026amp; Verspaget, M. (2020). Celebrity vs. Influencer endorsements in advertising: The role of identification, credibility, and Product-Endorser fit. International Journal of Advertising, 39(2), 258\u0026ndash;281. https://doi.org/10.1080/02650487.2019.1634898\u003c/li\u003e\n \u003cli\u003eTaboada, M., \u0026amp; Mann, W. C. (2006). Applications of Rhetorical Structure Theory. Discourse Studies, 8(4), 567\u0026ndash;588. https://doi.org/10.1177/1461445606064836\u003c/li\u003e\n \u003cli\u003eTomaselli, G., \u0026amp; Melia, M. (2014). THE ROLE OF INTERACTIVE TECHNOLOGIES FOR CSR COMMUNICATION. 8.\u003c/li\u003e\n \u003cli\u003eWang, C. L. (2021). New frontiers and future directions in interactive marketing: Inaugural Editorial. Journal of Research in Interactive Marketing, 15(1), 1\u0026ndash;9. https://doi.org/10.1108/JRIM-03-2021-270\u003c/li\u003e\n \u003cli\u003eWang, H.-J. (2019). Green city branding: Perceptions of multiple stakeholders. Journal of Product \u0026amp; Brand Management, 28(3), 376\u0026ndash;390. https://doi.org/10.1108/JPBM-07-2018-1933\u003c/li\u003e\n \u003cli\u003eXie, Q., \u0026amp; Wang, T. (Grace). (2022). Promoting corporate social responsibility message in COVID-19 advertising: How threat persuasion affects consumer responses to altruistic versus strategic CSR. Journal of Business Research, 148, 315\u0026ndash;324. https://doi.org/10.1016/j.jbusres.2022.04.073\u003c/li\u003e\n \u003cli\u003eXu, J., Guo, D., Xu, J., \u0026amp; Luo, C. (2023). How Do Multiple Actors Conduct Science Communication About Omicron on Weibo: A Mixed-Method Study. Media and Communication, 11(1), 306\u0026ndash;322. https://doi.org/10.17645/mac.v11i1.6122\u003c/li\u003e\n \u003cli\u003eYu, S., Eisenman, D., \u0026amp; Han, Z. (2021). Temporal Dynamics of Public Emotions During the COVID-19 Pandemic at the Epicenter of the Outbreak: Sentiment Analysis of Weibo Posts From Wuhan. Journal of Medical Internet Research, 23(3), e27078. https://doi.org/10.2196/27078\u003c/li\u003e\n \u003cli\u003eZhang, Z., Feng, G., Xu, J., Zhang, Y., Li, J., Huang, J., Akinwunmi, B., Zhang, C. J. P., \u0026amp; Ming, W. (2021). The Impact of Public Health Events on COVID-19 Vaccine Hesitancy on Chinese Social Media: National Infoveillance Study. JMIR Public Health and Surveillance, 7(11), e32936. https://doi.org/10.2196/32936\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"BERTopic, SnowNLP, CSR Advertising, Rhetorical Strategies","lastPublishedDoi":"10.21203/rs.3.rs-5697919/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5697919/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eIn the post-pandemic era, the relationship between Corporate Social Responsibility (CSR) advertising and consumer emotions has become a significant research focus within the CSR field. This study employs the BERTopic model and SnowNLP sentiment analysis, utilizing the Python 3.10 system to mine data from Weibo, to deeply explore how rhetorical themes in text affect the relationship between CSR advertising and consumer emotions. The findings reveal that: (1) CSR advertising has shifted from declarative expressions to pun-based rhetorical themes, with puns becoming the primary communication strategy; (2) Influenced by the COVID-19 pandemic, the context created by CSR advertising has gradually shifted from urging immediate consumption behaviors in the \"present\" to advocating that current actions can have a positive impact on the \"future\"; (3) Consumers show a strong enthusiasm and interest in CSR advertisements that use puns for marketing communication, far exceeding that for other rhetorical strategies.\u003c/p\u003e","manuscriptTitle":"The Evolution of Consumer Emotions in Response to Rhetorical Strategies in Corporate Social Responsibility Advertising: Evidence from Chinese Social Networks Before and After the COVID-19 Pandemic","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-01-09 14:07:52","doi":"10.21203/rs.3.rs-5697919/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":"760bda02-8e1d-4c9c-b1af-024d4f59aabe","owner":[],"postedDate":"January 9th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":42488583,"name":"Business and commerce/Business and management"},{"id":42488584,"name":"Humanities/Cultural and media studies"}],"tags":[],"updatedAt":"2025-03-28T15:08:47+00:00","versionOfRecord":[],"versionCreatedAt":"2025-01-09 14:07:52","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5697919","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5697919","identity":"rs-5697919","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}