AI Communication Coach

This project aims to develop a platform that utilizes Artificial Intelligence (AI) to assist users in improving their communication skills for public speaking and interviews. The coach can leverage various techniques, including:

• Speech Recognition: Capture user input through speech.
• Text Analysis: Analyze the text content to understand sentiment, keywords, and named entities.
• Pre-trained LLM Integration: Utilize a pre-trained Large Language Model (LLM) like GPT-j to generate informative responses.

Index

1. Features (In Progress)
2. Tech Stack
3. Getting Started
   • Prerequisites
   • Installation
   • Usage
4. Current Functionality
5. Future Development
6. Transformers and LLMs
7. License

Features (In Progress)

• Speech Recognition: Capture and analyze user speech for content and delivery.
• Text Analysis: Process user-provided text (e.g., job descriptions, presentation outlines) to understand context and generate relevant questions.
• Mock Interview Simulation: Simulate interview scenarios with AI-powered questions based on user input.
• Performance Analysis: Provide feedback on speech patterns, body language (future implementation), and content organization.

Tech Stack

• Python: Programming Language
• Jupyter Notebook: Development environment
• SpeechRecognition: Speech Recognition
• NLP Libraries: NLTK, spaCy, TextBlob
• Transformers: Hugging Face Transformers for text generation
• (Optional) OpenCV: Computer Vision for future features

Getting Started

Prerequisites

• Python 3.x (Ensure you have the latest version)
• Jupyter Notebook (Consider using Anaconda for a bundled environment)
• Additional libraries (install using pip in your terminal or the Jupyter Notebook environment):
  • transformers
  • speech_recognition
  • nltk
  • textblob
  • spacy
  • torch (for PyTorch backend)

Installation

1. Clone this repository :

   git clone https://github.com/JermaineV/AI_Coach
   cd AI_Coach

2. Install the required libraries:

   pip install -r requirements.txt # To install the dependencies, you can simply run this line
   # OR individual installations
   pip install transformers speech_recognition nltk textblob spacy torch

3. Download the necessary NLTK resources:

   import nltk
   nltk.download('punkt')

4. (Optional) Install PyAudio for speech recognition:

   • Download The necessary PyAudio file:

     pip install PyAudio

Usage

1. Open the Jupyter Notebook:

   jupyter notebook AI_coach.ipynb

2. Ensure you have a working internet connection for initial model downloads (if applicable).

3. Run the notebook cells (usually by pressing Shift + Enter) to execute the code.

Current Functionality

• Speech Recognition: Capture and analyze user speech.
• Text Analysis: Perform sentiment analysis, keyword extraction, and named entity recognition.
• LLM Integration: Use pre-trained LLMs to generate responses based on user input.
• Video Showcasing Current progress Video link

Future Development

• Enhance response generation using the LLM with context awareness and conversation history.
• Integrate feedback mechanisms to improve the coach's responses over time.
• Explore additional features like voice synthesis for coach responses or sentiment visualization.

Transformers and LLMs

The project utilizes the transformers library to interact with pre-trained LLMs.

Transformers Library

• Transformers Library: https://huggingface.co/transformers
• Install using pip install transformers

Pre-trained LLM

The code snippet utilizes the EleutherAI/gpt-j-6B model. Explore the Hugging Face Model Hub for various LLMs: https://huggingface.co/models

Downloading the LLM

There are two main approaches to using pre-trained LLMs with the transformers library:

A. Using Transformers Pipeline

• Pros:
  • Simpler setup, no need to manage model files.
• Cons:
  • Requires an internet connection during script execution.
  • Might have limitations on model customization or fine-tuning.

B. Downloading Model Weights

• Pros:
  • Offline functionality after initial download.
  • More flexibility for fine-tuning or advanced usage.
• Cons:
  • Requires additional storage space for the model files.
  • Downloading large models can take time.

Downloading Instructions

For option A (using pipeline), no additional downloads are necessary.

For option B (downloading weights), refer to the specific LLM's documentation on the Hugging Face Model Hub. Some models might provide pre-built transformers compatible weights files, while others may require specific download steps.

Important Note: Downloading large LLMs can require significant storage space and processing power. Consider your computational resources and the specific needs of your project before choosing a model.

License

This project is licensed under the MIT License. See the LICENSE file for more details.