Learn More from Less: Improving Conversational Recommender Systems via Contextual and Time-Aware Modeling

Anonymous

Abstract: Conversational Recommender Systems (CRS) aims to perform recommendations through interactive conversations. Prior work on CRS tends to incorporate more external knowledge to enhance performance. Given the fact that too much extra knowledge introduces the difficulty to balance among them and degenerates the generalizability, we propose to fully discover and extract the internal knowledge from the context. We capture both entity-level and contextual-level representations to jointly model user preferences for the recommendation, where a time-aware attention is designed to emphasize the recently appeared items in entity-level representations. We further use the pre-trained BART to initialize the generation module to alleviate the data scarcity and enhance the context modeling. Experiments on two public CRS datasets show that our model achieves comparable performance with less external knowledge and generalizes well to other domains. Further analyses demonstrate the effectiveness of our model in different scenarios.