HireSenseATS: AI-Powered Resume Screening and Recommendation Analysis

Introduction

Satya is an innovative AI-driven solution designed to revolutionize the HR landscape by streamlining the resume screening process and analyzing recommendation networks. This project aims to create a more efficient, reliable, and objective approach to talent acquisition in a world where careers are shaped by networks, connections, and credibility.

Table of Contents

1. Project Overview
2. Features
3. Installation
4. Usage
5. Project Structure
6. Methodology
7. Results
8. Future Directions

Project Overview

Satya employs advanced natural language processing (NLP) techniques, graph theory, and machine learning algorithms to:

1. Screen and score resumes
2. Analyze recommendation networks
3. Detect potential fraud in recommendations
4. Provide comprehensive candidate evaluations

The project is divided into three main parts:

1. Resume Screening and Initial Scoring
2. Network Analysis of Recommendations
3. Final Reliability Score Calculation

Features

• Extraction and analysis of key resume attributes
• Graph-based analysis of recommendation networks
• Sentiment analysis of recommendation letters
• Fraud detection in recommendations
• Composite scoring system for candidate evaluation
• Visualization of score distributions and network graphs

Installation

git clone /~https://github.com/vedprakashnautiyal/HireSenseATS.git
cd HireSenseATS
pip install -r requirements.txt

Usage

This will perform resume screening, recommendation analysis, final score calculation, and dashboard creation.

Methodology

Part 1: Resume Screening and Initial Scoring

This stage employs advanced NLP techniques to extract and analyze various measures from resumes:

• Years of Experience: Extracted using regex to identify date patterns.
• Education Level: Categorized using pattern matching algorithms.
• Spell Check Ratio: Implemented using the LanguageTool library.
• Resume Section Score: Semantic analysis to identify key resume sections.
• Brevity Score: Evaluates resume conciseness based on word count and content density.
• Skill Count and Relevance: Uses keyword extraction and semantic similarity analysis.
• Technical Score: Composite score derived from education, experience, and technical skills.
• Managerial Score: Uses sentiment analysis and achievement quantification algorithms.
• Overall Score: Weighted combination of all previous scores.
• Job Match Score: TF-IDF vectorization and cosine similarity with job description.

Part 2: Network Analysis of Recommendations

This stage constructs and analyzes a directed graph G(V,E) representing the recommendation network:

• PageRank: Identifies influential candidates and recommenders.
• Betweenness Centrality: Flags candidates with potential but not heavily networked.
• In-Degree: Measures the number of incoming recommendations.
• Reciprocal Recommendations: Detects and flags mutual recommendations.
• Cycle Detection: Identifies closed loops in the recommendation network.

Composite Credibility Score (CreditScore) Calculation:

CreditScore = IncomingEdgeWeightSum * ((0.4 * PageRank + 0.3 * Inverse_Betweenness) / (InDegree + 1)) - 0.3 * Reciprocity

Part 3: Final Reliability Score

• Combines scores from Parts 1 and 2
• Incorporates sentiment analysis of recommendation letters using LLaMA 3.1 model
• Generates a final normalized reliability score for each candidate

Results

The project outputs:

• CSV files with processed data and scores
• Visualizations of score distributions
• Network graphs showing recommendation patterns
• A final ranked list of candidates based on reliability scores
• An interactive dashboard for HR decision support

Future Directions

• Enhance recommendation credibility by matching company names and work terms
• Implement continuous learning to improve fraud detection accuracy

Application/UI

Features

• Candidate database browsing with pagination
• Skill-based filtering
• Detailed candidate profile view
• AI-generated resume summaries
• Visual representation of candidate scores
• Chatbot for asking questions about candidates

Prerequisites

• Python 3.7+
• Streamlit
• Pandas
• Groq API key
• Plotly
• LangChain

Installation

1. Install the required packages:

   pip install -r requirements.txt
   

2. Set up your API key:

   • Create a .env file in the project root
   • Add your Groq API key: API_KEY=your_api_key_here

Usage

1. Ensure you have the necessary CSV files in the project directory:

   • streamlit_table.csv
   • resume_text.csv

2. Run the Streamlit app:

   streamlit run app.py
   

3. Open your web browser and navigate to the provided local URL (usually http://localhost:8501)

How It Works

1. Candidate Database: Browse through candidate profiles with pagination.
2. Skill Filtering: Use the multiselect feature to filter candidates based on skills.
3. Profile View: Click on "View Profile" to see detailed information about a candidate.
4. Resume Summary: AI-generated summary of the candidate's resume.
5. Score Visualization: Circular bar charts display various scores for the candidate.
6. Chatbot: Ask questions about the candidate and receive AI-powered responses.