Project Title

Predicting Restaurant Review Ratings Using NLP and BERT

Problem Statement:

The goal of this project is to predict restaurant review ratings (1-5 stars) based on the text provided in user reviews. By leveraging advanced Natural Language Processing (NLP) techniques and the BERT model, the application can analyze the sentiment and context of a review and accurately predict the corresponding star rating. This solution can help restaurants and businesses quickly assess customer feedback and take action to improve customer satisfaction. The system provides real-time predictions and is designed to be both efficient and scalable for practical applications.

Features:

• Star Rating Prediction (1-5 Stars): The application predicts a restaurant review's star rating based on the sentiment and context of the review text.
• Real-Time Predictions: Users can input a review, and the model will provide an instant rating prediction.
• BERT for Sequence Classification: Utilizes the BERT (Bidirectional Encoder Representations from Transformers) model for advanced natural language understanding and accurate sentiment analysis.
• Preprocessing and Tokenization: The review text is tokenized and preprocessed, including padding and truncation, to fit BERT's input requirements.
• Flask Web Interface: A simple and interactive web application where users can submit reviews and receive predictions.
• PyTorch for Model Inference: Powered by PyTorch to handle the model's inference efficiently.
• Scalable and Extendable: The design allows for easy scaling or adaptation to other sentiment-based prediction tasks in various domains.

🔗 Links

🛠 Skills

Machine Learning, Deep Learning , NLP

NOTE

Before you clone this repo make sure you run the ipynb file and after getting the trained files then clone the repo and place the trained files in bert_sentiment_model folder

Installation

  pip install Flask
  pip install torch
  pip install transformers

Learnings and Challenges

1. Learned: Gained experience with BERT for NLP tasks and integrating machine learning models with Flask for real-time predictions.
2. Challenge: Ensuring correct model and tokenizer loading; resolved by validating checkpoints and following library documentation.
3. Challenge: Handling input preprocessing and padding for BERT; implemented a robust tokenization function.
4. Challenge: Efficient model inference without gradients; used torch.no_grad() for optimization.
5. Challenge: Creating a user-friendly web interface; designed intuitive forms and displayed predictions clearly.

Optimizations

NLP Model Choice: Used BERT for superior sentiment analysis accuracy over traditional algorithms.

Preprocessing: Implemented streamlined tokenization and padding to ensure consistency and compatibility with BERT.

Run Locally

Clone the project

  git clone https://github.com/AICraftAlchemy/Restaurent-Reviews-Ratings

Go to the project directory

  cd Restaurent-Reviews-Ratings

Install dependencies

  pip install -r requirements.txt

Start the server in terminal

  python app.py

Screenshots

Below are screenshots demonstrating the project:

These images showcase the functionality and user interface of the application.

Support

For support, email [email protected]

Feedback

If you have any feedback, please reach out to us at [email protected]

Phone: 7661081043