Multi-criteria Collaborative Recommendation System With Self- Attention : Study Of Impact Of Self-attention In Recommendation Phase

Résumé: Recommender systems have been very useful tools in different applications such as online marketing, e-commercial services, and social networking applications. They are information filtering technologies used to recommend products to users using specific techniques. The most common ones are collaborative filtering techniques. They are usually used on single rating datasets; however, they have been extended to work on multi-criteria datasets, which have been proved more accurate. Deep learning has achieved impressive results in many domains such as natural language processing Natural Language Processing (NLP). Recently, deep learning for recommender systems started to receive great interest, and there are many proposed models based on deep learning. Self-Attention is one of the deep learning methods that was first used in NLP domain, and then it was adapted in sequential recommender systems. However, as far as we know, there is not yet any study, which gathers multi-criteria recommendation and collaborative filtering with Self-Attention. In this work, we propose a novel multi-criteria collaborative filtering model with Self-Attention. Our experimental evaluation show a side-by-side comparison between multi-criteria collaborative filtering model with and without the use of Self-Attention. Results of our experiments show the success of using Self-Attention in multi-criteria collaborative filtering recommender systems.

Publié dans la revue: