Skip to content

This project is about sentiment analysis on Code-Mix data using LLM.

License

Notifications You must be signed in to change notification settings

skfrost19/Code-Mix-Data-Sentiment-Analysis

Repository files navigation

📌 Code-Mix Data Sentiment Analysis

👉 Introduction

This project is about sentiment analysis on Code-Mix data using Large Language Model (LLM). Code-Mixing is the phenomenon of alternating between two or more languages in a single conversation. This project aims to analyze the sentiment (positive, negative) of such data.

This repository uses LLM named Openchat - 3.5, 4-bit quantized for sentiment analysis.

Demo

Demo Image

🛠️ Setup and Installation

  1. Clone the repository: git clone https://github.com/skfrost19/Code-Mix-Data-Sentiment-Analysis.git
  2. Navigate into the project directory: cd Code-Mix-Data-Sentiment-Analysis
  3. Install the required dependencies: pip install -r requirements.txt

OR

  1. On linux system, run the setup.sh file to install the required dependencies and start the server.

  2. On windows system, run the setup.ps1 on powershell to install the required dependencies and start the server.

💻 Usage

To run the sentiment analysis, use the following command:

  • Run the fastAPI server: uvicorn webui:app --reload
  • Navigate to http://localhost:8000 in your browser to use the web interface.

📂 Project Structure

The project is structured as follows:

Code-Mix-Data-Sentiment-Analysis
├── config
│   ├── init_config.py
│   ├── __init__.py
│   ├── model_config.py
│   └── paths.toml
├── core
│   ├── exception.py
│   ├── __init__.py
│   └── logger.py
├── LICENSE.txt
├── README.md
├── requirements.txt
├── settings
│   ├── __init__.py
│   └── template.py
├── setup.ps1
├── setup.sh
├── static
│   └── images
│       └── favicon.png
├── structure
├── templates
│   └── index.html
├── tests
│   └── __init__.py
├── utils
│   ├── get_model.py
│   ├── pipeline.py
│   └── predictor.py
└── webui.py

📚 References & Credits

  • [1]: Openchat
  • [2]: TheBloke - For Quantized models
  • [3]: Huggingface
  • [4]: Langchain