Sentiment Analysis for movie reviews

Overview

This project is designed to perform sentiment analysis on textual data. It includes a machine learning model for predicting sentiment, a web interface for user interaction, and a backend to handle prediction requests.

The demo app is running on HuggingFace Spaces - https://huggingface.co/spaces/lyrisha/sentiment-analysis

Detailed description of this project, including accuracy results on tested datasets can be found on my blog

The purpose

• Show how to use HuggingFace pre-trained model and Transformers library to fine-tune it to the specific task of sentiment analysis.
• Show how to integrate a sentiment analysis model with a Gradio web application.

Files Description

src.training

• model_trainer.py: Contains the ModelTrainer class that loads user specified base model and trains the model with user specified dataset. The trained model is saved into HuggingFace model library.

src.prediction

• model_predictor.py: Contains the ModelPredictor class that loads a pre-trained sentiment analysis model and provides a method for making predictions on new text data.

notebooks

• example how to use pre-trained model from Hugging Face
• example how to make inference on custom fine-tuned model

demo

• app.py: The runner for demo. This script initializes the web application with Gradio library.
• ui.py: Handles the user interface aspects of the application. Generating and managing the HTML content returned to the user.

Setup and Installation

1. Clone the repository to your local machine.

git clone git@github.com:kanovotn/sentiment-analysis-model-trainer.git

2. Ensure you have Python 3.6+ installed.
3. Install the required dependencies:

pip install -r requirements.txt

Usage - Fine-tuning the model

For example how to fine-tune sentiment classificator refer to example notebook for fine-tunning

Usage - Evaluate the model

For example how to evaluate your transformer based model refer to notebook for inference and evaluation