Semantic Mapping of AI-Related Job Competencies in Construction: A Knowledge Graph Analysis

Overview

This repository contains the complete codebase and methodology for the research paper "Semantic Mapping of AI-Related Job Competencies in Construction: A Knowledge Graph Analysis" by Dr. M. Reza Hosseini from The University of Melbourne.

Key Contributions

• Novel Methodology: First systematic application of knowledge graph methodology to AI workforce analysis in construction
• Automated Pipeline: LLM-powered knowledge extraction from job postings data
• Network Analysis: Comprehensive analysis of AI role interconnections and competency patterns
• Practical Insights: Evidence-based guidance for workforce development and career pathways

Quick Start

Option 1: Google Colab (Recommended)

Option 2: Local Setup

# Clone the repository
git clone https://github.com/morehosseini/knowledge-graph-llms.git
cd knowledge-graph-llms

# Create conda environment
conda env create -f environment.yml
conda activate knowledge-graph-llms

# Or use pip
pip install -r requirements.txt

# Start Jupyter
jupyter notebook notebooks/

Repository Structure

├── data/                     # Sample and processed data
├── notebooks/               # Jupyter notebooks for analysis pipeline
├── src/                     # Source code modules
│   ├── data_processing/     # Data preprocessing utilities
│   ├── knowledge_extraction/ # LLM-based extraction tools
│   ├── graph_analysis/      # Network analysis functions
│   └── visualization/       # Plotting and visualization tools
├── results/                 # Generated figures and metrics
├── docs/                    # Documentation and guides
└── paper/                   # Paper manuscript and supplementary materials

Key Findings

🔍 87.5% of AI roles exist in an interconnected competency network
📊 Data Engineer emerges as the key bridge role (not Data Scientist)
💬 Communication skills are more central than technical skills for career mobility
📈 Construction shows 12.3% annual growth in AI recruitment (4x other industries)

Methodology

1. Data Collection: 283 AI-related job postings from major construction companies (Australia/New Zealand, 2021–2025)
2. Knowledge Extraction: Automated LLM-based entity and relationship extraction
3. Graph Construction: NetworkX-based knowledge graph with 2,362 nodes and systematic refinement
4. Network Analysis: Centrality metrics, clustering analysis, and pathway identification

Data Usage & Privacy

• Raw job posting data not included due to licensing restrictions
• Sample synthetic data provided for demonstration
• Original Lightcast data available with appropriate licensing
• All personal information anonymized

Citation

@article{hosseini2025semantic,
  title={Semantic Mapping of AI-Related Job Competencies in Construction: A Knowledge Graph Analysis},
  author={Hosseini, M. Reza},
  institution={The University of Melbourne},
  year={2025},
  doi={10.xxxx/xxxxxx}
}

License

This project is licensed under the MIT License - see the LICENSE file for details.

Contact

Dr. M. Reza Hosseini
Senior Lecturer in Construction Technology
Faculty of Architecture, Building and Planning
The University of Melbourne
📧 mreza.hosseini@unimelb.edu.au
🌐University Profile

Acknowledgments

• Lightcast for providing labor market analytics data
• OpenAI for LLM capabilities
• The University of Melbourne for research support

Warning: The notebooks and graph files here may not reflect the final, revised versions used in the study.