Bridging Conversational and Collaborative Signals for Conversational Recommendation
This addresses the issue of interaction sparsity in conversational recommendation for users, though it is incremental as it builds on existing methods by integrating collaborative signals.
The paper tackles the problem of conversational recommendation systems struggling due to lack of collaborative filtering signals by introducing a dataset linking Reddit conversations with MovieLens interactions and proposing an LLM-based framework to align recommendations with CF embeddings, resulting in a 12.32% increase in Hit Rate and 9.9% improvement in NDCG.
Conversational recommendation systems (CRS) leverage contextual information from conversations to generate recommendations but often struggle due to a lack of collaborative filtering (CF) signals, which capture user-item interaction patterns essential for accurate recommendations. We introduce Reddit-ML32M, a dataset that links Reddit conversations with interactions on MovieLens 32M, to enrich item representations by leveraging collaborative knowledge and addressing interaction sparsity in conversational datasets. We propose an LLM-based framework that uses Reddit-ML32M to align LLM-generated recommendations with CF embeddings, refining rankings for better performance. We evaluate our framework against three sets of baselines: CF-based recommenders using only interactions from CRS tasks, traditional CRS models, and LLM-based methods relying on conversational context without item representations. Our approach achieves consistent improvements, including a 12.32% increase in Hit Rate and a 9.9% improvement in NDCG, outperforming the best-performing baseline that relies on conversational context but lacks collaborative item representations.