AI-Powered-HR-Assistant

Faster, Fairer, and More Efficient Hiring Using Graph RAG

Graph-RAG is a sophisticated question-answering system that leverages knowledge graphs and Retrieval-Augmented Generation (RAG) to provide accurate and context-aware responses. The system processes PDF documents, extracts structured information, and stores it in a Neo4j graph database for efficient querying and retrieval.

Chatbot UI:

Link to connect with us - https://bit.ly/poster-informs-2025

Features

• PDF document processing and text extraction
• Automatic knowledge graph construction from documents
• Structured information extraction
• Graph-based question answering
• Interview question generation
• Interactive web interface using Streamlit
• Support for multiple LLM providers (Groq, OpenAI, Ollama)
• Flexible embedding options (Cohere, OpenAI, Ollama)

Prerequisites

• Python 3.8+
• Neo4j Database
• API keys for:
  • Groq (or other LLM provider)
  • Cohere (or other embedding provider)

Installation

1. Clone the repository:

git clone https://github.com/mahithabsl/Graph-RAG.git
cd Graph-RAG

2. Install required dependencies:

pip install -r requirements.txt

3. Configure Neo4j:

   • Install and start Neo4j database
   • Update the config.ini file with your Neo4j credentials

4. Set up API keys:

   • Update the config.ini file with your API keys for the LLM and embedding providers

Configuration

The config.ini file contains all necessary configuration settings:

• Neo4j connection details
• LLM provider settings
• PDF processing parameters
• Embedding model configuration

Project Structure

• Pdf_2_Text.py: Handles PDF text extraction and resume parsing
• MetaData_Extraction.ipynb: Extracts structured metadata from parsed text
• Pdf_2_Graph.py: Constructs knowledge graph from extracted information
• Graph_QA.py: Implements the question-answering system
• Generate_Interview_Questions.ipynb: Generates interview questions based on resume content
• data/: Directory for storing PDF documents
• static/: Static assets for the web interface

Usage

Step 1: Resume Parsing and PDF to Text

from graph_rag.Pdf_2_Text import process_pdf

# Process a PDF resume
text_content = process_pdf("path/to/your/resume.pdf")

Step 2: Metadata Extraction

# Run the MetaData_Extraction.ipynb notebook
# This extracts structured information from the parsed text

Step 3: PDF to Graph Construction

from graph_rag.Pdf_2_Graph import process_document

# Process a document and build the knowledge graph
process_document("path/to/your/document.pdf", metadata_info={}, meta={})

Step 4: Graph-based Question Answering

streamlit run graph_rag/Graph_QA.py

Access the web interface at http://localhost:8501

Step 5: Interview Question Generation

# Run the Generate_Interview_Questions.ipynb notebook
# This generates relevant interview questions based on the resume content

How It Works

1. Resume Parsing and Text Extraction:

   • PDFs are processed and converted to text
   • Structured information is extracted from resumes

2. Metadata Extraction:

   • Key information is identified and structured
   • Education, experience, skills, and other relevant data are extracted

3. Knowledge Graph Construction:

   • Entities and relationships are identified
   • Information is stored in Neo4j graph database
   • Graph structure enables complex relationship queries

4. Question Answering:

   • User questions are processed
   • Relevant information is retrieved from the graph
   • LLM generates context-aware responses

5. Interview Question Generation:

   • Resume content is analyzed
   • Relevant interview questions are generated
   • Questions are tailored to the candidate's background

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.