User review ratings are crucial for assessing the quality of restaurants, movies, products, and services online. Traditionally, predicting these ratings involved analyzing the content of user reviews using natural language processing techniques. While this approach allows for the consideration of diverse data by aggregating reviews from all users, it's hindered by the fact that users have unique tendencies and expressions, resulting in varying ratings for the same review. To address this challenge, a novel method is proposed in this paper: a BiLSTM model that incorporates the user's average rating alongside their review content. By doing so, this method aims to enhance prediction accuracy by accounting for individual user characteristics. This approach considers the textual content of reviews and integrates the user's typical rating behavior, resulting in more personalized and precise rating predictions.