Premier League AI Chatbot

An AI-powered chatbot for querying Premier League football data, built as coursework for ISYS30221 Artificial Intelligence at Nottingham Trent University (2025).

The chatbot combines multiple AI techniques — AIML pattern matching, logic-based reasoning, fuzzy logic, CNN image classification, and a live football API — to answer a wide range of questions about the Premier League.

---

Features

Core Tasks

• AIML Chatbot (Task a) — natural language Q&A about the Premier League using a pattern-matched AIML knowledge base, with Wikipedia fallback for unknown queries
• Logic-based reasoning (Task b) — fact storage and retrieval for teams, managers, and players using CSV-backed knowledge bases; supports "I know that..." and "Check that..." commands
• CNN Image Classification (Task c) — classifies Premier League club logos from images using a trained and hyperparameter-tuned Convolutional Neural Network (Keras/TensorFlow)

Extra Tasks

• Live Football API — fetches real-time Premier League standings, top scorers, and upcoming matches from the football-data.org API
• Fuzzy Logic — estimates the likelihood of a team finishing in the top 4 using fuzzy membership functions
• Hyperparameter Optimisation — CNN model tuned using Keras Tuner for improved classification accuracy

---

Tech Stack

Component	Technology
Chatbot engine	AIML (aiml library)
Machine learning	TensorFlow / Keras
Hyperparameter tuning	Keras Tuner
Fuzzy logic	Custom implementation
Live data	football-data.org API
Fallback knowledge	Wikipedia API
Language	Python

---

Project Structure

Chatbox2/
├── main.py                   # Entry point — chatbot main loop
├── core/
│   ├── knowledge_base.py     # Fallback Q&A knowledge base
│   ├── team_logic.py         # Team fact storage and checking
│   ├── manager_logic.py      # Manager logic and queries
│   ├── players_logic.py      # Player team lookups
│   ├── my_fuzzy_logic.py     # Fuzzy top-4 likelihood
│   └── football_data_api.py  # Live Premier League API calls
├── models/
│   ├── cnn_train.py          # CNN training script
│   ├── cnn_tune.py           # Keras Tuner hyperparameter search
│   ├── class_names.py        # EPL club label names
│   └── epl_logos_cnn_tuned.h5 # Trained CNN model (not in repo — see below)
├── aiml_files/
│   └── mybot-basic.xml       # AIML pattern definitions
└── data/
    ├── QA_KB.csv             # Q&A knowledge base
    ├── logical-kb-2.csv      # Logic facts
    ├── manager_logic_kb.csv  # Manager data
    ├── players_logic_kb.csv  # Player data
    ├── fuzzy_facts.csv       # Fuzzy logic data
    └── epl-logos-big/        # EPL logo images for CNN training

---

Getting Started

Prerequisites

• Python 3.9+
• A football-data.org API key (free tier available)

Installation

1. Clone the repository:

   git clone https://github.com/mkinteh/YOUR_REPO_NAME.git
   cd Chatbox2
   

2. Create and activate a virtual environment:

   python -m venv .venv
   source .venv/bin/activate  # On Windows: .venv\Scripts\activate
   

3. Install dependencies:

   pip install aiml wikipedia tensorflow keras-tuner requests python-dotenv
   

4. Create a .env file in the root directory:

   FOOTBALL_DATA_API_KEY=your_api_key_here
   

5. Run the chatbot:

   python main.py
   

CNN Model

The trained model file (epl_logos_cnn_tuned.h5) is not included in the repository due to its file size. To regenerate it:

python models/cnn_train.py   # Train the base model
python models/cnn_tune.py    # Run hyperparameter tuning

---

Example Commands

> Who manages Arsenal?
> I know that Manchester City is a top club
> Check that Manchester City is a top club
> Show me the Premier League standings
> Show me the top scorers
> How likely is Chelsea to finish in the top 4?
> Identify logo   (then provide an image path)

---

Author

Muhamed Kinteh (N0994178)
BSc Computer Science & Mathematics — Nottingham Trent University, 2025
Module: ISYS30221 Artificial Intelligence