Publication

Abstract : In the current era of excessive information, recommendation systems are essential for assisting users to discover relevant content efficiently, ensuring high levels of user satisfaction and engagement. The proposed work focuses on enhancing content-based collaborative filtering using advanced word embed ding techniques like Word2Vec, GloVe, and BERT to capture semantic relationships between items. This study utilizes weighted word embeddings, derived from user-provided ratings, to enhance the recommendation system’s capability in identifying and predicting user preferences. By incorporating these weighted embeddings, our approach seeks to improve the precision and relevance of recommendations, eventually increasing user satisfaction. The study conducts a comparative analysis of these word embedding techniques based on cosine similarity scores. Among the three weighted embeddings, GloVe shows the highest cosine similarity score for selected users UserId-68 and UserId-122.