Predicting Personalized Textual Reviews via Collaborative Filtering using Document Embedding
Young Park, Harika Bollam · 2023
As the number of online product reviews is exploding in an e-commerce website, it is increasingly difficult for users to view all reviews and find their most helpful reviews for a product that they are interested in. Providing personalized helpful reviews on new products to individual users is valuable for users to decide on purchasing new products. Predicting the expected review on a new product by a user even before the user purchases will provide useful insight to the user on how the user would like the product after purchase. We propose a novel personalized textual product review prediction method via collaborative filtering. The method is based on user similarity using product ratings and semantic textual similarity via document embedding of textual product reviews. Initial experiments of the proposed prediction method on a textual product review dataset show promising results. The textual reviews predicted based on semantic textual similarity via document embedding are semantically comparable with the actual textual review.