This project focused on developing a neural network-based classifier for IMDB movie reviews, with the goal of comparing the performance of various model architectures. We utilized Pandas for data processing, GloVe embeddings for word representations, and implemented the models using Keras. While a basic shallow neural network showed limited performance, replacing the dense layers with unidirectional and bidirectional LSTMs led to substantial improvements in both accuracy and F1 scores—especially with the bidirectional LSTM, which yielded the best results. This work was completed as part of the CSE440: NLP II course.