A Hybrid Recommendation System Based on Bidirectional Encoder Representations

Abstract

Using auxiliary data about items provides more accurate item recommendations when utilizing deep learning in the recommendation system. Users often read item descriptions during online shopping, which contain key information about the item and its features. However the item descriptions are in unstructured form and using them in the deep learning model is a problem. In this study, we integrate a pioneering Natural Language Processing technique into a recommendation system to create an item embedding vector from unstructured item description text. The experimental results show that the proposed approach is efficient in generating more accurate recommendations by creating item embedding vectors from unstructured item description text.