Sentiment Analysis using BERT and Hugging Face Transformers

Overview

This project aims to perform sentiment analysis using the BERT model and Transformers by Hugging Face. The project includes the following key components:

• Scraping data from Google Play
• Preprocessing the data
• Building and training a sentiment classifier
• Creating a REST API for sentiment analysis

Setup and Installation

1. Clone the repository:

   git clone https://github.com/CruiseDevice/Sentiment-Analysis-using-BERT.git
   cd sentiment_analyzer

2. Install the required dependencies:

   pip install -r requirements.txt

3. Download and prepare the pre-trained model: Ensure the best_model_state.bin is placed in the ./models directory or adjust the path in config.json.

Running the API

To run the API, use the following command:

uvicorn sentiment_analyzer.api:app --reload

This will start the server on http://localhost:8000.

Testing the API Endpoint

You can test the API endpoint using the following URL: http://localhost:8000/predict`

Request Body

To predict the sentiment of a text, send a POST request with the following JSON body:

{
    "text": "This app is a total waste of time!"
}

Sanple output

The API will respond with a JSON object containing the predicted sentiment, the confidence score, and the probabilities for each sentiment class:

{
    "sentiment": "negative",
    "confidence": 0.9952511787414551,
    "probabilities": {
        "negative": 0.9952511787414551,
        "neutral": 0.0025495770387351513,
        "positive": 0.002199336187914014
    }
}