An exploration of recommender systems explanation paradigms: generating and evaluating syntactic, semantic, and generative models with knowledge graphs: An Extended Abstract

• André Levi Zanon University College Cork
• Leonardo C. Dutra da Rocha UFSJ
• Marcelo Garcia Manzato USP

DOI: https://doi.org/10.5753/webmedia_estendido.2025.16351

Abstract

Recommender Systems (RS) generate suggestions for users based on their past interactions and the interactions of other users. However, because RS rely on identifying similarities between users, they may expose users to a limited range of content. In addition, as algorithms have become increasingly complex, users often cannot understand why a recommendation is generated. In this thesis, we explored three approaches—syntactic, semantic, and generative—to generate explanations for recommendations and investigate their potential impact on the diversity of recommendations, with the goal of broadening user interests. Our results show that explanations can play an important role in increasing item diversity, thereby helping mitigate filter bubble effects. We also observed an improvement in the explanation quality across the approaches, from syntactic to semantic and finally to generative models.

Keywords: Recommender Systems, Explanation, Explanation Evaluation

References

Giacomo Balloccu, Ludovico Boratto, Gianni Fenu, and Mirko Marras. 2022. Post Processing Recommender Systems with Knowledge Graphs for Recency, Popularity, and Diversity of Explanations. In Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval (Madrid, Spain) (SIGIR ’22). Association for Computing Machinery, New York, NY, USA, 646–656. DOI: 10.1145/3477495.3532041

Nava Tintarev and Judith Masthoff. 2007. A Survey of Explanations in Recommender Systems. DOI: 10.1109/icdew.2007.4401070

André Zanon, Leonardo Rocha, and Marcelo Manzato. 2024. O Impacto de Estratégias de Embeddings de Grafos na Explicabilidade de Sistemas de Recomendação. In Proceedings of the 30th Brazilian Symposium on Multimedia and the Web (Juiz de Fora/MG). SBC, Porto Alegre, RS, Brasil, 231–239. DOI: 10.5753/webmedia.2024.241857

André Levi Zanon, Leonardo Chaves Dutra da Rocha, and Marcelo Garcia Manzato. 2022. Balancing the trade-off between accuracy and diversity in recommender systems with personalized explanations based on Linked Open Data. Knowledge-Based Systems 252 (2022), 109333. DOI: 10.1016/j.knosys.2022.109333

André Levi Zanon, Leonardo Chaves Dutra Da Rocha, and Marcelo Garcia Manzato. 2024. Model-Agnostic Knowledge Graph Embedding Explanations for Recommender Systems. Communications in Computer and Information Science, 3–27. DOI: 10.1007/978-3-031-63797-1_1

André L. Zanon, Luan Souza, Diany Pressato, and Marcelo G. Manzato. 2020. WordRecommender: An Explainable Content-Based Algorithm based on Sentiment Analysis and Semantic Similarity. In Proceedings of the Brazilian Symposium on Multimedia and the Web (São Luís, Brazil) (WebMedia ’20). Association for Computing Machinery, NewYork, NY, USA, 181–184. DOI: 10.1145/3428658.3431093

André Levi Zanon, Luan Souza, Diany Pressato, and Marcelo Garcia Manzato. 2022. A user study with aspect-based sentiment analysis for similarity of items in content-based recommendations. Expert Systems 39, 8 (2022). DOI: 10.1111/exsy.12991