Sentiment deep learning algorithm for multi-criteria recommendation
Lamia Berkani, Mohamed Zaouidi, Rodaina Brahimi · 2022
Multi-criteria recommender systems have emerged as an important trend in studying recommendation systems. They outperform and have been proved to be more precise than single-criterion systems. On the other hand, deep learning (DL) techniques have shown promising results in various fields. Recently, several studies on single-criterion recommender systems explored DL and sentiment analysis. We propose in this article a novel sentiment DL based algorithm for multi-criteria recommendation using autoencoders and sentiment information. Two sentiment analysis models have been employed, LSTM and LSTM with Word2Vec. Experiments on the TripAdvisor multi-criteria dataset demonstrated the contribution of sentiment information and that our approach outperformed other exiting methods.