Neural Microblogging Followee Recommender System Based on Pre-trained Transformer, and Topic Models

Microblogging platforms have experienced tremendous success, allowing users to publish their posts and follow microbloggers having the same interests. In recent years, the followee recommendation task has got more attention. In fact, the daily huge amount of information produced by users makes the task of finding like-minded ones very hard and time-consuming. Thus, the need for a system to assist users in finding such information becomes primordial. In this paper, the authors propose a followee recommender system to assist users in finding like-minded micro-bloggers to follow. Moreover, our approach is based on the ELECTRA model for language modeling, LDA for topic modeling, and LSTM model for classification. To show the effectiveness of our approach, the authors conduct extensive experiments based on a real dataset harvested from Twitter. Experiment results show that our approach outperforms state-of-art approaches in both precision and recall measures.