RAAC LLM Analysis Code Package

Associated Research Paper

This code package is the official implementation of the research paper:

Title: "Leveraging Large Language Models to Classify and Inspect Defects in Reinforced Autoclaved Aerated Concrete (RAAC)"

Status: Under peer review

Authors: Liu, L. et al.

Year: 2025

Overview

This repository contains the full Python codebase supporting our research paper on RAAC defect analysis using LLMs. The code implements a structured, LLM-based workflow to:

• Extract and compare RAAC definitions
• Check for explicit RAAC mentions
• Extract defect-related information using a 7-question prompt framework
• Combine results into a consolidated dataset for analysis

The scripts were developed using Claude 3 Opus API and tested on a dataset of UK-based documents.

Features

• Intelligent definition extraction and comparison
• Structured defect analysis framework
• Comprehensive data aggregation
• Parallel prompt evaluation system

Project Structure

Core Files

1. Definition Extraction

2. Explicit Mention Check

3. Defect Analysis

Installation

1. Clone the repository:

   git clone https://github.com/lixuliu/raac-llm-analysis-code-package.git
   cd raac-llm-analysis-code-package

2. Create a virtual environment (recommended):

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

4. Set up your API key:

   export ANTHROPIC_API_KEY='your-api-key-here'  # On Windows: set ANTHROPIC_API_KEY=your-api-key-here

Usage

1. Set up your environment:

   # Set your Anthropic API key
   export ANTHROPIC_API_KEY='your-api-key-here'  # On Windows: set ANTHROPIC_API_KEY=your-api-key-here
   
   # Install required dependencies
   pip install -r requirements.txt

2. Prepare your data:

   • Create a folder containing your PDF documents to analyze
   • Ensure all PDFs are readable and not corrupted

3. Run the analysis pipeline:

   # Step 1a: Extract RAAC definitions
   python "01-Prompt Setting 1-a_RAAC_DefinitionExtraction.py"
   
   # Step 1b: Check for RAAC definition presence
   python "01-Prompt Setting 1-b_RAAC_DefinitionPresenceCheck_BinaryFlag.py"
   
   # Step 2a: Check for explicit RAAC mentions
   python "02_Prompt Setting 2-a_RAAC_ExplicitMentionCheck.py"
   
   # Step 2b: Check for RAAC mention presence
   python "02_Prompt Setting 2-b_RAAC_ExplicitMentionCheck_BinaryFlag.py"
   
   # Step 3: Extract defect information
   python 03_RAAC_DefectExtraction_SevenQuestions.py
   
   # Step 4: Combine the results
   python 04_RAAC_CombineExtractedDefectData.py