Recommender System Using Transformer Model: A Systematic Literature Review

Abstract

Online transactions are significant in the pandemic era. Using online transactions can minimize the risk of physical contact with disease transmission between buyers and sellers. However, with so many choices of items, it becomes challenging for users to decide which item suits their needs. For this reason, the recommender system was created as a handy tool. Recommender systems can help provide ratings, compare with other user data, use personal transaction history, use current events, or combine the above methods. Currently, computer science experts are constantly trying to improve recommender systems. In 2017 a new method emerged that uses transformers as one of the deep learning models. The combination of recommender systems and transformers can process extensive data, create different weights for each input data, and process data without sequentially allowing parallel processing and reducing training time significantly. Many papers in various countries are continuously trying to improve this methodology. In this literature review, we try to analyze the technology used, the dataset used, and the area where the technology is implemented. In this case, we carry out collecting papers, then filtering, classifying and analyzing, and making conclusions.