This project demonstrates a sentiment analysis on Amazon reviews using the BERT model. It uses bert-tensorflow for text processing and TensorFlow Hub to fine-tune BERT embeddings for classification tasks. This notebook is designed to evaluate how well BERT can classify reviews as positive or negative based on the provided Amazon dataset.
- Data Preprocessing: Includes text cleaning and tokenization using BERT's
FullTokenizer. - Model Building: Constructs a BERT-based classification model using
tensorflow_hub. - Training and Evaluation: Splits the dataset, trains the model, and evaluates its performance on the test set.
- Visualization: Provides insights into model performance through visualizations.
To run this notebook, install the required packages:
pip install bert-tensorflow tqdm tensorflow_hub- Clone this repository and navigate to the project directory.
- Run the notebook using Jupyter or Google Colab
This project aims to analyze the effectiveness of BERT in sentiment classification on real-world data from Amazon reviews.