File size: 1,513 Bytes
bde793d | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 | # Text-Based Chatbot Performance Analysis
This project implements an end-to-end performance analysis system for text-based chatbots using Machine Learning. It features a Bidirectional LSTM model with an Attention mechanism to evaluate chatbot responses based on provided context and facts.
## Features
- **Exploratory Data Analysis**: Visual insights into engine distribution and performance correlations.
- **Advanced LSTM Model**: Uses Bidirectional LSTM and Attention layers for high-accuracy evaluation.
- **Context-Aware Prediction**: Evaluates responses not just on linguistics but also on factual consistency.
- **Modern Web Interface**: Glassmorphic UI with real-time performance analytics.
- **Flask Backend**: Robust API for model inference.
## Project Structure
- `train_model.py`: Training pipeline for the advanced model.
- `app.py`: Flask server for real-time predictions.
- `explore_data.py`: EDA script for dataset visualization.
- `BP_MHS_V1.csv`: The core dataset.
- `templates/` & `static/`: Frontend assets.
## How to Run
1. Install dependencies:
```bash
pip install -r requirements.txt
```
2. Train the model:
```bash
python train_model.py
```
3. Start the application:
```bash
python app.py
```
4. Access the UI at `http://127.0.0.1:5000`.
## Model Insights
The system uses an Attention mechanism to focus on critical parts of the facts and responses, ensuring the expert verdict is both accurate and contextually relevant.
|