DeepLTRS: A deep latent recommender system based on user ratings and reviews

Dingge Liang, Marco Corneli, Charles Bouveyron, Pierre Latouche

Abstract: Highlights•DeepLTRS provides high quality rating recommendations based on observed user ratings and texts of product reviews.•VAE architecture is used as a generative model for both the ordinal matrix and the document-term matrix.•A new row-column alternated mini batch strategy is designed to deal with user-product dependencies.•Numerical experiments on simulated and real-world data sets demonstrate that deepLTRS outperforms the state-of-the-art.