{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,9]],"date-time":"2026-10-09T10:50:50Z","timestamp":1791543050521,"version":"4.3.4"},"reference-count":35,"publisher":"JMIR Publications Inc.","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["JMIR Infodemiology"],"abstract":"Abstract<\/jats:title>\n \n Background<\/jats:title>\n Public health agencies increasingly use social media platforms such as Twitter (subsequently rebranded X, X Corp) to monitor public responses to official health communications and support timely decision-making during health crises. Public replies to communications from official agencies provide valuable insights into population-level discourse and engagement with health policies. However, analyzing large-scale short-text responses remains challenging because social media replies are often brief, informal, ambiguous, and linguistically fragmented. Topic modeling offers a scalable approach for identifying thematic patterns in such data, but its performance is often limited when applied to social media short texts. Recent advances in large language models (LLMs) provide new opportunities to improve text normalization before topic modeling, but their value and limitations for public health infoveillance remain underexplored.<\/jats:p>\n <\/jats:sec>\n \n Objective<\/jats:title>\n This study develops and evaluates TM-Rephrase, a model-agnostic LLM-based rephrasing framework designed to improve the performance of topic models for short public health social media texts while ensuring rephrasing preserves the original semantic meaning, stance or intent, and tone.<\/jats:p>\n <\/jats:sec>\n \n Methods<\/jats:title>\n We analyzed 25,027 public replies to official Centers for Disease Control and Prevention tweets on Twitter collected between May 2020 and November 2022. TM-Rephrase transformed informal and noisy short texts into more standardized expressions using 2 prompt-guided schemes: general rephrasing and colloquial-to-formal rephrasing. Original and rephrased texts were analyzed using multiple topic models, with rephrasing generated by Gemini 2.5 Flash, GPT\u20114o mini, and Mistral-7B-Instruct. Topic quality was evaluated using coherence, uniqueness, redundancy, and diversity. We also conducted an expert semantic-fidelity validation study in which 4 expert raters evaluated whether rephrased texts preserved the original meaning, stance or intent, and tone.<\/jats:p>\n <\/jats:sec>\n \n Results<\/jats:title>\n \n TM-Rephrase generally improved topic quality across models, rephrasing schemes, and LLM backbones. For latent Dirichlet allocation,\n \n \n \n \n C<\/mml:mi>\n <\/mml:mrow>\n \n v<\/mml:mi>\n <\/mml:mrow>\n <\/mml:msub>\n <\/mml:math>\n <\/jats:inline-formula>\n coherence increased from 0.3094 without rephrasing to 0.5004 with colloquial-to-formal rephrasing. For BERTopic (BERT-based topic modeling),\n \n \n \n \n C<\/mml:mi>\n <\/mml:mrow>\n \n v<\/mml:mi>\n <\/mml:mrow>\n <\/mml:msub>\n <\/mml:math>\n <\/jats:inline-formula>\n coherence improved from 0.4078 to 0.4734. Diversity-related metrics also improved in most rephrasing settings. Additional checks using\n \n \n \n \n C<\/mml:mi>\n <\/mml:mrow>\n \n N<\/mml:mi>\n P<\/mml:mi>\n M<\/mml:mi>\n I<\/mml:mi>\n <\/mml:mrow>\n <\/mml:msub>\n <\/mml:math>\n <\/jats:inline-formula>\n ,\n \n \n \n \n C<\/mml:mi>\n <\/mml:mrow>\n \n U<\/mml:mi>\n C<\/mml:mi>\n I<\/mml:mi>\n <\/mml:mrow>\n <\/mml:msub>\n <\/mml:math>\n <\/jats:inline-formula>\n , and\n \n \n \n \n U<\/mml:mi>\n <\/mml:mrow>\n \n M<\/mml:mi>\n A<\/mml:mi>\n S<\/mml:mi>\n S<\/mml:mi>\n <\/mml:mrow>\n <\/mml:msub>\n <\/mml:math>\n <\/jats:inline-formula>\n generally supported the main improvement pattern. Sensitivity analysis across different topic numbers further suggested that the coherence gains were not limited to the main setting. Expert validation showed favorable semantic preservation overall, with a mean rating of 3.90 (SD 1.01) and median of 4.00 (IQR 3.00-5.00) of 5.00, and acceptable interrater reliability, intraclass correlation coefficient =0.736. General rephrasing received stronger semantic-fidelity ratings than colloquial-to-formal rephrasing, suggesting that stronger formalization may improve topic readability while introducing greater risk of changes to the original meaning, stance or intent, and tone.\n <\/jats:p>\n <\/jats:sec>\n \n Conclusions<\/jats:title>\n TM-Rephrase can enhance the coherence, distinctiveness, and interpretability of topic modeling outputs for short, noisy public health social media data. However, it is a text-normalization preprocessing strategy for topic enhancement rather than a fully neutral replacement for original public discourse.<\/jats:p>\n <\/jats:sec>","DOI":"10.2196\/98319","type":"journal-article","created":{"date-parts":[[2026,8,31]],"date-time":"2026-08-31T09:00:07Z","timestamp":1788166807000},"page":"e98319-e98319","source":"Crossref","is-referenced-by-count":0,"title":["Identifying Public Response Topics to the Centers for Disease Control and Prevention\u2019s COVID-19 Communications on Social Media: Infoveillance Study Using Large Language Model\u2013Based Rephrasing"],"prefix":"10.2196","volume":"6","author":[{"ORCID":"https:\/\/orcid.org\/0009-0009-5073-097X","authenticated-orcid":false,"given":"Wangjiaxuan","family":"Xin","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5735-5804","authenticated-orcid":false,"given":"Shuhua","family":"Yin","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2419-0131","authenticated-orcid":false,"given":"Albert","family":"Park","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-9576-0293","authenticated-orcid":false,"given":"Yaorong","family":"Ge","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2316-111X","authenticated-orcid":false,"given":"Shi","family":"Chen","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1010","published-online":{"date-parts":[[2026,10,8]]},"reference":[{"issue":"2","key":"R1","doi-asserted-by":"publisher","first-page":"129","DOI":"10.1080\/13183222.2021.1923622","article-title":"Public sphere in crisis mode: how the COVID-19 pandemic influenced public discourse and user behaviour in the Swiss Twitter-sphere","volume":"28","author":"Rauchfleisch","journal-title":"Javnost"},{"issue":"3","key":"R2","doi-asserted-by":"publisher","DOI":"10.2196\/26482","article-title":"Understanding concerns, sentiments, and disparities among population groups during the COVID-19 pandemic via Twitter data mining: large-scale cross-sectional study","volume":"23","author":"Zhang","journal-title":"J Med Internet Res"},{"issue":"1","key":"R3","doi-asserted-by":"publisher","DOI":"10.2196\/49756","article-title":"Dynamic associations between Centers for Disease Control and Prevention social media contents and epidemic measures during COVID-19: infoveillance study","volume":"4","author":"Yin","journal-title":"JMIR Infodemiology"},{"key":"R4","unstructured":"Timeline of X. Wikipedia. URL: https:\/\/en.wikipedia.org\/wiki\/Timeline_of_Twitter [Accessed 10-09-2026]"},{"key":"R5","doi-asserted-by":"publisher","DOI":"10.1007\/s10462-023-10471-x","article-title":"A systematic review of the use of topic models for short text social media analysis","volume":"56","author":"Laureate","journal-title":"Artif Intell Rev"},{"key":"R6","volume":"3","author":"Blei","journal-title":"J Mach Learn Res"},{"key":"R7","doi-asserted-by":"crossref","unstructured":"Hong L Davison BD . Empirical study of topic modeling in Twitter. Presented at: Proceedings of the First Workshop on Social Media Analytics; Jul 25-28, 2010:80-88; Washington, DC. [doi: 10.1145\/1964858.1964870]","DOI":"10.1145\/1964858.1964870"},{"key":"R8","doi-asserted-by":"crossref","unstructured":"Rajagopal D Olsher D Cambria E Kwok K . Commonsense-based topic modeling. Presented at: Proceedings of the second international workshop on issues of sentiment discovery and opinion mining; Aug 11-12, 2013:1-8; Chicago, IL. [doi: 10.1145\/2502069.2502075]","DOI":"10.1145\/2502069.2502075"},{"key":"R9","unstructured":"BERTopic. URL: https:\/\/maartengr.github.io\/BERTopic\/ [Accessed 10-09-2026]"},{"key":"R10","unstructured":"Angelov D . Top2Vec: distributed representations of topics. arXiv. Preprint posted online on Aug 19, 2020. [doi: 10.48550\/arXiv.2008.09470]"},{"key":"R11","doi-asserted-by":"crossref","unstructured":"Wu X Nguyen T Zhang DC Wang WY Luu AT . FASTopic: pretrained transformer is a fast, adaptive, stable, and transferable topic model. Presented at: Advances in Neural Information Processing Systems 37; Dec 10-15, 2024:84447-84481; Vancouver, BC, Canada. [doi: 10.52202\/079017-2683]","DOI":"10.52202\/079017-2683"},{"key":"R12","doi-asserted-by":"crossref","unstructured":"Wu X Luu AT Dong X . Mitigating data sparsity for short text topic modeling by topic-semantic contrastive learning. Presented at: Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing; Dec 7-11, 2022:2748-2760; Abu Dhabi, United Arab Emirates. [doi: 10.18653\/v1\/2022.emnlp-main.176]","DOI":"10.18653\/v1\/2022.emnlp-main.176"},{"issue":"1","key":"R13","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1080\/01463373.2021.1995772","article-title":"Information seeking and information sharing during the COVID-19 pandemic","volume":"70","author":"Yang","journal-title":"Commun Q"},{"key":"R14","doi-asserted-by":"publisher","DOI":"10.2196\/59425","article-title":"Long COVID Discourse in Canada, the United States, and Europe: topic modeling and sentiment analysis of Twitter data","volume":"26","author":"AbuRaed","journal-title":"J Med Internet Res"},{"issue":"4","key":"R15","doi-asserted-by":"publisher","first-page":"563","DOI":"10.17269\/s41997-023-00783-4","article-title":"Exploring communication by public health leaders and organizations during the pandemic: a content analysis of COVID-related tweets","volume":"114","author":"James","journal-title":"Can J Public Health"},{"issue":"2","key":"R16","doi-asserted-by":"publisher","DOI":"10.2196\/19273","article-title":"Tracking Social Media Discourse About the COVID-19 Pandemic: Development of a Public Coronavirus Twitter Data Set","volume":"6","author":"Chen","journal-title":"JMIR Public Health Surveill"},{"key":"R17","doi-asserted-by":"crossref","unstructured":"Maini P Seto S Bai R Grangier D Zhang Y Jaitly N . Rephrasing the web: a recipe for compute and data-efficient language modeling. Presented at: Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1; 14044-14072; Bangkok, Thailand. [doi: 10.18653\/v1\/2024.acl-long.757]","DOI":"10.18653\/v1\/2024.acl-long.757"},{"key":"R18","doi-asserted-by":"publisher","DOI":"10.1016\/j.neucom.2025.131928","article-title":"A comprehensive survey on automatic text summarization with exploration of LLM-based methods","volume":"663","author":"Zhang","journal-title":"Neurocomputing"},{"issue":"1","key":"R19","doi-asserted-by":"publisher","DOI":"10.1038\/s41598-024-60210-7","article-title":"A multimodal approach to cross-lingual sentiment analysis with ensemble of transformer and LLM","volume":"14","author":"Miah","journal-title":"Sci Rep"},{"key":"R20","doi-asserted-by":"crossref","unstructured":"Liu Z Zhou Y Zhu Y et al. Information retrieval meets large language models. Presented at: Companion Proceedings of the ACM Web Conference 2024; May 13-17, 2024:1586-1589; Singapore, Singapore. [doi: 10.1145\/3589335.3641299]","DOI":"10.1145\/3589335.3641299"},{"key":"R21","unstructured":"Gemini 25 flash model. Google Cloud. URL: https:\/\/docs.cloud.google.com\/vertex-ai\/generative-ai\/docs\/models\/gemini\/2-5-flash [Accessed 10-09-2026]"},{"key":"R22","unstructured":"GPT-4o mini: advancing cost-efficient intelligence. OpenAI. URL: https:\/\/openai.com\/index\/gpt-4o-mini-advancing-cost-efficient-intelligence\/ [Accessed 10-09-2026]"},{"key":"R23","unstructured":"Mistral-7B-instruct-v02. Hugging Face. URL: https:\/\/huggingface.co\/mistralai\/Mistral-7B-Instruct-v0.2 [Accessed 10-09-2026]"},{"key":"R24","doi-asserted-by":"crossref","unstructured":"Renze M . The effect of sampling temperature on problem solving in large language models. Presented at: Findings of the Association for Computational Linguistics: EMNLP 2024; Nov 12-16, 2024:7346-7356; Miami, FL. [doi: 10.18653\/v1\/2024.findings-emnlp.432]","DOI":"10.18653\/v1\/2024.findings-emnlp.432"},{"key":"R25","doi-asserted-by":"crossref","unstructured":"Reimers N Gurevych I . Sentence-BERT: sentence embeddings using siamese BERT-networks. Presented at: Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP); 3982-3992; Hong Kong, China. [doi: 10.18653\/v1\/D19-1410]","DOI":"10.18653\/v1\/D19-1410"},{"key":"R26","doi-asserted-by":"crossref","unstructured":"R\u00f6der M Both A Hinneburg A . Exploring the space of topic coherence measures. Presented at: Proceedings of the Eighth ACM International Conference on Web Search and Data Mining; Feb 2-6, 2015; Shanghai, China. [doi: 10.1145\/2684822.2685324]","DOI":"10.1145\/2684822.2685324"},{"key":"R27","unstructured":"Wu X . Towards effective neural topic modeling. Nanyang Technological University; 2024. URL: https:\/\/dr.ntu.edu.sg\/bitstreams\/2b7465e1-7444-4cda-9078-9e7163cd4006\/download [Accessed 10-09-2026]"},{"key":"R28","unstructured":"English Wikipedia database dump. Wikimedia Foundation. URL: https:\/\/dumps.wikimedia.org\/enwiki\/latest\/ [Accessed 10-09-2026]"},{"key":"R29","doi-asserted-by":"crossref","unstructured":"Nan F Ding R Nallapati R Xiang B . Topic modeling with Wasserstein autoencoders. Presented at: Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics; Jul 28 to Aug 2, 2019:6345-6381; Florence, Italy. [doi: 10.18653\/v1\/P19-1640]","DOI":"10.18653\/v1\/P19-1640"},{"issue":"131","key":"R30","first-page":"1","volume":"20","author":"Burkhardt","journal-title":"J Mach Learn Res"},{"key":"R31","doi-asserted-by":"publisher","DOI":"10.1162\/tacl_a_00325","article-title":"Topic modeling in embedding spaces","volume":"8","author":"Dieng","journal-title":"Trans Assoc Comput Linguist"},{"key":"R32","unstructured":"Newman D Lau JH Grieser K . Automatic evaluation of topic coherence. Presented at: Human Language Technologies: The 2010 Annual Conference of the North American Chapter of the Association for Computational Linguistics; Jun 2-4, 2010; Los Angeles, CA. URL: https:\/\/aclanthology.org\/N10-1012.pdf [Accessed 10-09-2026]"},{"key":"R33","unstructured":"Mimno D Wallach H Talley E . Optimizing semantic coherence in topic models. Presented at: Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing; Jul 27-31, 2011; Edinburgh, Scotland, UK. [doi: 10.5555\/2145432.2145462]"},{"key":"R34","unstructured":"Lau GR Low WY Koh SM Nah FFH Hartanto A . Evaluating AI alignment in LLMs: output analysis of value priorities across 75 models with human benchmarking. arXiv. Preprint posted online on May 16, 2026. [doi: 10.48550\/arXiv.2506.12617]"},{"issue":"3","key":"R35","doi-asserted-by":"publisher","DOI":"10.1080\/08989621.2025.2544331","article-title":"GAIDeT (Generative AI Delegation Taxonomy): A taxonomy for humans to delegate tasks to generative artificial intelligence in scientific research and publishing","volume":"33","author":"Suchikova","journal-title":"Account Res"}],"container-title":["JMIR Infodemiology"],"original-title":[],"language":"en","deposited":{"date-parts":[[2026,10,9]],"date-time":"2026-10-09T10:00:44Z","timestamp":1791540044000},"score":1,"resource":{"primary":{"URL":"https:\/\/infodemiology.jmir.org\/2026\/1\/e98319"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,10,8]]},"references-count":35,"URL":"https:\/\/doi.org\/10.2196\/98319","relation":{},"ISSN":["2564-1891"],"issn-type":[{"value":"2564-1891","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,10,8]]},"article-number":"v6i2e98319"}}