Topic Discovery and Hotspot Analysis of Sentiment Analysis of Chinese Text Using Information-Theoretic Method.

Area of Science:

• Computer Science
• Statistical Science
• Natural Language Processing

Background:

• Sentiment analysis is a rapidly evolving research area with significant implications across various scientific disciplines.
• Understanding research trends is crucial for scholars to navigate the dynamic landscape of text sentiment analysis literature.

Purpose of the Study:

• To propose and validate a new model for automated topic discovery and trend analysis in text sentiment analysis literature.
• To identify key research themes, their evolution, and emerging challenges within the field from 2012 to 2022.

Main Methods:

• Utilized FastText for keyword vectorization and cosine similarity for synonym merging.
• Applied hierarchical clustering with Jaccard coefficient for topic categorization and information gain for characteristic word extraction.
• Conducted time series analysis and constructed a four-quadrant matrix to visualize topic distribution and research trends across different time phases.

Main Results:

• Identified 12 distinct research categories within text sentiment analysis literature from 2012-2022.
• Online opinion analysis of social media (e.g., microblogs) emerged as a prominent current research focus.
• Observed significant shifts in research focus between the 2012-2016 and 2017-2022 periods.

Conclusions:

• The proposed model effectively identifies and analyzes research trends in text sentiment analysis.
• Future research should prioritize integrating sentiment lexicons, traditional machine learning, and deep learning methods.
• Addressing semantic disambiguation in aspect-level sentiment analysis and advancing multimodal/cross-modal sentiment analysis are critical future directions.