This document presents a hybrid model for sentiment-based rating prediction that combines BERT (Bidirectional Encoder Representations from Transformers) and LSTM (Long Short-Term Memory) networks to enhance accuracy and understanding of sentiment in text data. The proposed system addresses limitations of traditional and existing methods by effectively capturing contextual and sequential information, minimizing manual feature engineering, and employing techniques to prevent overfitting. Preliminary results from experiments on various datasets indicate that the hybrid approach significantly outperforms existing models in terms of robustness and accuracy.