Knowledge-Enhanced Conversational Recommendation via Transformer-based Sequential Modelling

TL;DR

This work addresses the challenge of modeling sequential dependencies among items and item-related entities discussed in conversational recommender systems. It introduces TSCR, a Transformer-based model with a Cloze objective to capture bidirectional sequence dynamics, and TSCRKG, an enhanced version that uses knowledge graphs for offline embedding initialization and multi-hop path augmentation. Empirical results on ReDial and TG-ReDial show that TSCR outperforms strong baselines, while TSCRKG consistently yields further gains, demonstrating the value of integrating structured knowledge with sequential modeling. The approach offers improved recommendation quality and better potential for explainability through KG-driven paths, highlighting practical benefits for knowledge-rich CRSs and suggesting avenues for extending to response generation and sentiment-aware reasoning.

Abstract

In conversational recommender systems (CRSs), conversations usually involve a set of items and item-related entities or attributes, e.g., director is a related entity of a movie. These items and item-related entities are often mentioned along the development of a dialog, leading to potential sequential dependencies among them. However, most of existing CRSs neglect these potential sequential dependencies. In this article, we first propose a Transformer-based sequential conversational recommendation method, named TSCR, to model the sequential dependencies in the conversations to improve CRS. In TSCR, we represent conversations by items and the item-related entities, and construct user sequences to discover user preferences by considering both the mentioned items and item-related entities. Based on the constructed sequences, we deploy a Cloze task to predict the recommended items along a sequence. Meanwhile, in certain domains, knowledge graphs formed by the items and their related entities are readily available, which provide various different kinds of associations among them. Given that TSCR does not benefit from such knowledge graphs, we then propose a knowledge graph enhanced version of TSCR, called TSCRKG. In specific, we leverage the knowledge graph to offline initialize our model TSCRKG, and augment the user sequence of conversations (i.e., sequence of the mentioned items and item-related entities in the conversation) with multi-hop paths in the knowledge graph. Experimental results demonstrate that our TSCR model significantly outperforms state-of-the-art baselines, and the enhanced version TSCRKG further improves recommendation performance on top of TSCR.

Figures (5)

• Figure 1: An example dialog from the ReDial dataset. The mentioned items (i.e., movies) are highlighted in blue color, and item-related entities in red color.
• Figure 2: The overview of our model. Our model extracts the items and item-related entities to form an input sequence, and then uses the knowledge graph and bidirectional Transformer to generate item recommendations.
• Figure 3: The performance of TSCR and TSCRKG on ReDial with the ordinal number of item predictions.
• Figure 4: The performance of TSCR and TSCRKG on TG-ReDial with the ordinal number of item predictions.
• Figure 5: Effect of hidden dimensionality and mask proportion.