Based on the LDA topic model and linear regression method,the analysis and mining of the character-istics of the scientific popularization content on the

Qiongtong FANG; Xinrong WU; Yishen CHEN; Jiabi LIANG; Huibin ZHAO; Wenji LUO; Meirong ZHANG

Return

Based on the LDA topic model and linear regression method,the analysis and mining of the character-istics of the scientific popularization content on the"PSM medicine shield public welfare"official ac-count were conducted

VernacularTitle:基于LDA主题模型和线性回归方法对"PSM药盾公益"公众号科普作品内容特点的分析与挖掘
Author: Qiongtong FANG ¹ ; Xinrong WU ; Yishen CHEN ; Jiabi LIANG ; Huibin ZHAO ; Wenji LUO ; Meirong ZHANG
Author Information

1. 中山大学附属第五医院广东珠海 519000
Publication Type:Journal Article
Keywords: PSM medicine shield public welfare; Science communication; Text mining; LDA model; Linear regression
From: Modern Hospital 2025;25(9):1428-1432,1437
CountryChina
Language:Chinese
Abstract: Objective Text mining of content characteristics of original science popularization articles on the WeChat public account"PSM Medicine Shield Public Welfare"based on LDA model and Linear Regression.Methods Through web crawling techniques,we collected 4,292 original pharmaceutical science popularization articles and associated comment data from the"PSM Drug Shield Public Welfare"WeChat Official Account to analyze the content distribution patterns and characteristics of pharmaceutical science communication.Employing Latent Dirichlet Allocation(LDA)modeling,we systematically categorized and mined article themes to identify public demand for science literacy and explore strategic directions for precision-targeted phar-maceutical science dissemination.Results The analysis of 4 292 original science popularization articles from the"PSM Medi-cine Shield Public Welfare"account showed an average readership of 1 815.73±4 385.31.Articles in headline positions,pub-lished on weekends,or using exclamatory titles achieved higher readership.Most articles(1 000-2 000 words)used direct-open-ing title strategies.Traditional Chinese medicines had the highest readership among drug categories.Top content categories in-cluded drug monographs,medicinal diets,disease medication guidance,myth clarification,and disease science.Linear regres-sion analysis identified headline placement,word count,title phrasing,title strategy,target audience,drug category,and content type as potential factors influencing readership.The LDA model with 9 s revealed key themes:management of drug adverse reac-tions,dermatological medication dosing,anti-infection effects of medicinal diets,and pediatric vaccination/health monitoring.Conclusion The"PSM Medicine Shield Public Welfare"WeChat account primarily disseminates pharmaceutical science content focusing on adverse drug reactions,pediatric medication safety,TCM-based health preservation,disease treatment protocols,symptom recognition guidelines,dosage optimization,and toxicity management.Pharmaceutical professionals should prioritize content length control,evidence-based title strategies,and thematic alignment with public health priorities during science commu-nication content creation.