The proliferation of recommender systems (RS) research has attracted researchers to explore new tools and techniques to address several issues in the domain. In this context, text mining has received considerable acceptance for solving cold start and data sparsity problems in recommender systems. Additionally, text mining has been exploited to design modern recommender systems through sentiment analysis of user reviews, social network posts, web data, etc. These sentiment identifications help create user profiles, identify preferences, understand the context in which a review is provided, and extract features to assess item quality, facilitating better recommendations. This paper presents a comprehensive and systematic review of text mining-based recommender systems (TMRS). We analyze how text mining techniques have been integrated into different RS paradigms and propose a novel taxonomy that classifies TMRS based on underlying text mining approaches and recommendation approaches. In addition, we examine commonly used evaluation metrics and discuss how they are applied to assess TMRS performance. Key research challenges and open issues are identified, along with promising future research directions. This review provides a structured overview of the state of the art in TMRS and serves as a useful reference for researchers and practitioners seeking to design, evaluate, and advance next-generation recommender systems.
Anwar et al. (2026) studied this question.