Intent Classifier using Machine Learning

This repository contains a Python-based intent classification system that leverages Natural Language Processing (NLP) techniques to identify user intents from text inputs. The implementation includes training a Support Vector Machine (SVM) model on an expanded dataset and provides a command-line interface for testing intent classification.

Features

• Intent Detection: Determines the user's intent with a confidence score.
• Dataset Expansion: Automatically triples the provided dataset to enhance training.
• Customizable Threshold: Allows setting a confidence threshold for intent prediction.
• Real-Time User Interaction: Accepts user input for on-the-fly intent classification.

Requirements

• Python 3.x
• Required libraries (install via pip):
  • scikit-learn
  • numpy

How It Works

1. Dataset Preparation: A predefined dataset of intents and examples is expanded to improve model performance.
2. Model Training: A pipeline consisting of a TF-IDF vectorizer and an SVM classifier is trained on the dataset.
3. Intent Classification: The model predicts the intent of user inputs based on the highest confidence score.

Usage

1. Clone the repository:

git clone https://github.com/Walapalam/intellihack_raccoons_task2.git
cd intellihack_raccoons_task2

2. Install dependencies:

pip install -r requirements.txt

3. Run the script:

python q2.py

4. Enter your message at the prompt to see the predicted intent and confidence score. Type quit to exit the program.

Example Interaction

Enter your message (or type 'quit' to exit): Hello
Predicted Intent: Greet, Confidence: 0.85

Enter your message (or type 'quit' to exit): What's the weather like today?
Predicted Intent: Inquiry, Confidence: 0.76

File Overview

• q2.py: Main script containing the dataset, model training, and classification logic.

Future Improvements

• Include additional intents and examples for better coverage.
• Enhance the fallback mechanism for handling low-confidence predictions.
• Provide a web or GUI-based interface for easier user interaction.