Deep Critiquing for VAE-based Recommender Systems

• DOI: 10.1145/3397271.3401091
• Publication Date: 2020-07-25T07:50:08Z
• Authors: Kai Luo; Hojin Yang; Ga Wu; Scott Sanner
• Keywords: Popularity; Autoencoder; Aside; Rank (graph theory)

Providing explanations for recommended items not only allows users to understand the reason receiving recommendations but also provides with an opportunity refine by critiquing undesired parts of explanation. While much research focuses on improving explanation recommendations, less effort has focused interactive recommendation allowing a user critique explanations. Aside from traditional constraint- and utility-based systems, end-to-end deep learning based approach in literature so far, CE-VNCF, suffers unstable inefficient training performance. In this paper, we propose Variational Autoencoder (VAE) system mitigate these issues improve overall The proposed model generates keyphrase-based generated their personalized recommendations. Our experiments show promising results: (1) is competitive terms general performance comparison state-of-the-art recommenders, despite having augmented loss function support critiquing. (2) can generate high-quality compared or item keyphrase popularity baselines. (3) more effective refining than where rank critiquing-affected drops while remains stable. summary, paper presents significantly improved method multi-step recommender systems VAE framework.