This project focuses on sentiment analysis of human-written movie reviews using deep learning models. The objective is to classify whether a given movie review expresses a positive or negative sentiment. Two main models were developed: one based on Bidirectional Long Short-Term Memory (BiLSTM) and another based on Convolutional Neural Networks (CNN).

Initially, simple versions of these models were established as baseline models. Subsequently, several modifications were experimented with and added on top of these baselines to improve the accuracy of sentiment classification for each review.

Baseline Model Establishment: Simple versions of BiLSTM and CNN models were developed to serve as baseline models for sentiment analysis.

Model Modifications: Various modifications were applied to the baseline models to enhance their performance. These modifications could include changes in architecture, hyperparameter tuning, or feature engineering.

Training and Evaluation: The modified models were trained on labeled movie review datasets and evaluated on separate test datasets to measure their performance in sentiment classification.

Experimentation: Throughout the development process, different configurations and techniques were experimented with to identify the most effective approaches for improving sentiment analysis accuracy.

The results of the experimentation phase, including the performance metrics of the baseline models and the modified models, will be presented in detail. This will include accuracy, precision, recall, and F1 score for each model variant.

The conclusions drawn from the experimental results, along with insights into the effectiveness of various modifications and approaches in improving sentiment analysis accuracy, will be provided. Recommendations for future research directions or potential areas for further optimization may also be discussed.