A Multifacet Hierarchical Sentiment-Topic Model with Application to Multi-Brand Online Review Analysis

2025-02-26

AI Summary

This paper addresses the challenge of brand-specific sentiment polarity detection in multi-brand online reviews. We propose a hierarchical topic-sentiment generative framework that jointly models textual content and numerical ratings. Our method introduces: (1) a novel hierarchical Polish urn mechanism that automatically disentangles general topics from brand-specific topics—the first such approach; and (2) a unified variational inference framework that jointly optimizes hierarchical topic distributions and fine-grained sentiment regression. Integrating Bayesian generative modeling, Pólya urn sampling, and sentiment analysis, the framework is validated on synthetic data and two real-world review datasets. Results demonstrate significant improvements in cross-brand sentiment ranking accuracy and topic interpretability. The approach provides an interpretable, scalable foundation for comparative market performance analysis across multiple brands.

Technology Category

• Machine Learning: Learning Preferences or Rankings
• Natural Language Processing: Sentiment Analysis, Stylistic Analysis, and Argument Mining
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Web Mining and Content Analysis: Sentiment analysis and opinion mining
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Multi-brand analysis based on review comments and ratings is a commonly used strategy to compare different brands in marketing. It can help consumers make more informed decisions and help marketers understand their brand's position in the market. In this work, we propose a multifacet hierarchical sentiment-topic model (MH-STM) to detect brand-associated sentiment polarities towards multiple comparative aspects from online customer reviews. The proposed method is built on a unified generative framework that explains review words with a hierarchical brand-associated topic model and the overall polarity score with a regression model on the empirical topic distribution. Moreover, a novel hierarchical Polya urn (HPU) scheme is proposed to enhance the topic-word association among topic hierarchy, such that the general topics shared by all brands are separated effectively from the unique topics specific to individual brands. The performance of the proposed method is evaluated on both synthetic data and two real-world review corpora. Experimental studies demonstrate that the proposed method can be effective in detecting reasonable topic hierarchy and deriving accurate brand-associated rankings on multi-aspects.

Problem

• Analyzes sentiment in multi-brand online reviews.
• Detects brand-associated sentiment polarities across aspects.
• Enhances topic-word association for brand-specific topics.

Innovation

• Hierarchical sentiment-topic model
• Hierarchical Polya urn scheme
• Brand-associated topic regression