Integrating BERTopic and Pareto-Based Topic Reduction for Sentiment-Frequency Topic Mapping
Year: 2026
Type: conference-paper
Abstract: The increasing availability of consumer reviews in the digital world has encouraged the use of sentiment analysis and topic modeling to extract insights from unstructured text data. This work proposes a framework that integrates BERTopic with Pareto-based topic reduction to form a sentiment-and frequency-based topic map, the Customer Voice Impact Matrix (CVIM). The map represents each topic in a two-dimensional space based on sentiment tendency and discussion intensity. Using six months of online game review data, the proposed approach applies Pareto analysis to retain dominant topics that cumulatively represent the majority of players’ attention, while filtering out topics with relatively low impact. The results show that Pareto-based topic reduction simplifies the topic structure and improves the clarity of the sentiment–frequency map, thus supporting a more focused and decision-oriented customer voice analysis.
Institutions: Binus University, Universitas Ekuitas Indonesia
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7204755170