Zhihu Urban Discourse on Suzhou (2015–2025): Annotated Sentiment and Topic Classification Dataset

This dataset was compiled for the study “Adding Sociodigital Discourse to Public Engagement Tools in the Design and Planning of Industrial Cities in China: An NLP-Based Sentiment Analysis in the Context of Suzhou.” It contains user-generated text responses from the Chinese Q&A platform Zhihu related to urban life in Suzhou between 2015 and 2025. The dataset was collected using a custom Python-based web scraping script (Selenium with ChromeDriver) applied to 22 Zhihu forums addressing themes such as housing, transport, governance, social integration, and cultural identity. Only publicly accessible comments were collected; no login, bypass, or private data was used. Each entry in the dataset includes: Comment Text (Chinese, UTF-8 encoded) Publication Year and Timestamp Engagement Metrics: upvotes, number of replies Sentiment Scores: polarity score (continuous range from –1 to +1), categorical sentiment (positive/neutral/negative) derived from the IDEA-CCNL/Erlangshen-Roberta-110M-Sentiment model Topic Labels: thematic classification into categories such as housing, transport, public space, retail, social belonging, exclusion/class, city identity, comparative judgement, governance, and decentralization Weighted Sentiment: polarity score multiplied by upvote count to capture resonance Size & Format ~6,000 comments CSV format, UTF-8 encoded Columns: 回答者姓名 (respondent name, anonymized), 发布或编辑日期 (date), 赞同数 (upvotes), 评论数 (number of replies), 回答内容 (answer text), 自动分类主题 (topic labels), 情绪倾向 (sentiment label), 正面概率 (positive probability), 负面概率 (negative probability), 情绪极性分数 (sentiment polarity score), 评论权重极性分数 (weighted sentiment score). Potential Uses The dataset can support research in: Urban planning and governance in China Computational social science and sentiment analysis Chinese language NLP and sarcasm detection Comparative studies of industrial cities and public engagement Ethical Statement This dataset is derived exclusively from publicly accessible Zhihu forums. To preserve user privacy, all personal identifiers have been removed. The dataset should be used strictly for academic and non-commercial purposes.