A Study on Movie Users’ Perception Dimensions and Sentiment Orientation Based on LDA Topic Modeling and SnowNLP Sentiment Analysis: Evidence from Douban Movie Reviews
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
With the sustained prosperity of the Chinese film market and the deepening development of internet social platforms, Douban Movie, as one of the most influential film review platforms in China, has accumulated a vast amount of user-generated content (UGC), serving as a critical data source for understanding audience film consumption behaviors and emotional attitudes. However, traditional manual analysis methods are limited in both efficiency and depth when confronted with massive volumes of unstructured review text. This study employs text mining techniques, integrating Latent Dirichlet Allocation (LDA) topic modeling with SnowNLP sentiment analysis, to systematically analyze 4,306 user reviews from the Douban Movie platform spanning 2020 to 2025. The original dataset encompasses reviews of 484 films (412 after preprocessing). Following preprocessing steps including text cleaning, jieba word segmentation, stopword filtering, and synonym normalization, the optimal number of topics was determined using the CV coherence metric, ultimately identifying five core thematic dimensions: Narrative Structure and Plot Evaluation, Acting Performance and Character Portrayal, Audio-Visual Effects and Production Quality, Social Value and Cultural Expression, and Viewing Experience and Platform Interaction. Sentiment analysis results indicate that audiences exhibit the highest emotional engagement with narrative structure and social value themes, with positive review proportions of 42.3% and 38.7%, respectively. The audio-visual effects theme shows the highest proportion of negative reviews at 58.2%, reflecting elevated audience expectations for the technical quality of domestic Chinese films. Annual trend analysis reveals fluctuating audience sentiment from 2020 to 2025, with a notable increase in review activity during 2023–2024. This study constructs a text mining analytical framework applicable to Chinese-language film reviews, providing data-driven empirical support for content creation and marketing decision-making in the film industry.
Fields of Study & Subjects
Sentiment Analysis and Opinion Mining
Cinema and Media Studies
Computational and Text Analysis Methods
Publication details
- Journal / source
- International Theory and Practice in Humanities and Social Sciences
- Volume
- 3
- Issue
- 2
- Pages
- 24–38
- Publication date
- 31 March 2026
- Language
- EN