Chatbot Sportif Expert

Overview

This project is a Retrieval-Augmented Generation (RAG) application designed to analyze sports performances and provide tactical insights. It utilizes a FAISS index for efficient retrieval and integrates a chatbot interface using Streamlit.

Dataset

The dataset was sourced from WhoScored, with a limited number of articles manually collected. Due to the constraints of local inference and the small dataset, the model's performance in text generation can sometimes be impacted.

Features

• Sports Performance Analysis: Provides insights based on retrieved articles.
• Chat Interface: Built with Streamlit for an interactive experience.
• FAISS Indexing: Enables fast and efficient retrieval of relevant information.
• Locally Hosted Model: Runs inference without relying on external APIs.

Limitations

• Limited Dataset: Only a few articles were manually taken from WhoScored.
• Local Inference Constraints: Running on local hardware affects performance and speed.
• Limited Model Parameters: Using a restricted set of parameters can lead to occasional inconsistencies in generated responses.

Installation

1. Clone the repository:

   git clone https://github.com/Ihssane5/chatbot-sportif
   cd chatbot-sportif

2. Install dependencies:

   pip install -r requirements.txt

3. Run the application:

   streamlit run app.py

Don't forget to generate your Hugging Face api token and place it in .env file

Demo

Watch a short demo of the chatbot in action:

Acknowledgments

• WhoScored for the dataset.
• FAISS for fast retrieval capabilities.
• Streamlit for the UI framework.

Feel free to contribute and improve the project! 🚀