Dual-space Hierarchical Learning for Goal-guided Conversational Recommendation

Anonymous

Abstract: Proactively and naturally guiding the dialog from the non-recommendation context~(e.g., Chit-chat) to the recommendation scenario is crucial for the Conversational Recommender System~(CRS). Prior studies mainly focus on planning the next dialog goal~(e.g., chat on a movie star) conditioned on the previous dialog. However, we find the dialog goals can be simultaneously observed at different levels, which can be utilized to improve CRS.In this paper, we propose the \textit{\textbf{D}ual-space \textbf{H}ierarchical \textbf{L}earning}~(\textbf{DHL}) to leverage multi-level goal sequences and their hierarchical relationships for conversational recommendation. Specifically, we exploit multi-level goal sequences from both the representation space and the optimization space. In the representation space, we propose the hierarchical representation learning where a cross attention module derives mutually enhanced multi-level goal representations. Additionally, we propose a soft labeling strategy to gradually guide the optimization direction. Experiments on two real-world datasets verify the effectiveness of our approach.