AI Assisted Analysis Tool

This project is an open-source, locally run AI-assisted text analysis tool powered by Ollama. It now supports two distinct workflows:

1. Analysis Workflow

Purpose: Analyze any tabular data (Excel or CSV) using AI, not limited to abstracts. This workflow is flexible and user-friendly, allowing you to select which columns to analyze and how the results are reported.

Key Features: - Works with any Excel or CSV file (not just abstracts) - Lists all columns and lets you select identifier and content columns by name - Lets you define what you want the AI to identify in your data (custom prompt) - Allows you to set the number of analysis runs for each row - Optionally aggregates AI responses for consensus and confidence - Optionally appends all reporting info (prompt, LLM, runs, hardware, consensus summary, etc.) to the bottom of the Excel output file - Enhanced reporting: includes prompt, LLM used, number of runs, analysis duration, CPU/GPU info - Cross-platform: Windows, macOS, Linux

How to Use: 1. Prepare your Excel or CSV file and place it in your chosen data input folder. 2. Make sure you also have a data output folder for your file to be saved in after the analysis. 3. Run the script: python ai_assisted_analysis.py 4. Follow the prompts: - Select the AI model (or press Enter for the recommended model) - Enter what you want the program to identify in your data (custom prompt) - Choose columns for identifier and content by name - Set the number of runs for analysis - Optionally aggregate AI responses for consensus and confidence - Optionally append all reporting info to the bottom of the output Excel file 5. Review your results in the output Excel file (includes consensus columns and reporting info if selected)

2. Zotero Abstracts Workflow

Purpose: Analyze bibliographic abstracts exported from Zotero. This workflow is designed for users working specifically with Zotero data and abstracts.

Key Features:

How to Use: 1. Export your collection from Zotero as a CSV or Excel file 2. Use the provided scripts (e.g., methods.py, results.py, location.py, theory.py, n_themes.py) in the python_for_Zotero_abstracts folder to run targeted analyses on your exported data 3. Follow the prompts in each script for results and consensus aggregation 4. Review your results in the output Excel file

Key Features

Requirements

Requirements

• ollama - Python client for Ollama
  • Ollama for Windows
  • Ollama for Mac
  • Ollama for Linux
    • pandas - Data manipulation and analysis
    • tqdm - Progress bars
    • openpyxl - Excel file handling

AI Assisted Analysis Tool

Project Goals

This project aims to provide tools for AI-assisted analysis and aggregation of responses, supporting research workflows and data exploration.

Features

• Aggregates AI responses from various sources
• Modular analysis methods (see other_analysis/ and python_for_Zotero_abstracts/)
• Extensible for new data formats and sources

Installation

1. Clone the repository: powershell git clone https://github.com/hleve/AI_Assisted_Analysis_Tool.git
2. Install dependencies: powershell pip install -r requirements.txt

Usage Example

Run the main analysis script: powershell python ai_assisted_analysis.py

Getting Started

See documentation.md for a step-by-step guide.

Contributing

Contributions are welcome! Please see CONTRIBUTING.md for contribution guidelines, issue templates, and the code of conduct.

License

See LICENSE for details.

---

Project Logic and Methodology

This tool is designed to leverage AI models for research analysis, with a focus on reliability and reproducibility. The core logic includes:

• Replication for Error Minimization:

  • Analyses are run multiple times (replicated) to reduce the impact of random errors or outlier responses from AI models.
  • Aggregation and consensus algorithms are used to combine results, improving reliability and confidence in findings.

• Prompt Engineering for Quantitative Analysis:

  • Prompts can be crafted to request numerical values from AI models (e.g., ratings, counts, scores).
  • These outputs can be collected across replications and subjected to standard statistical analyses, such as ANOVA or T-tests, to assess differences, trends, or significance.

This approach allows researchers to harness the flexibility of AI while maintaining scientific rigor and transparency in their workflows.

Prompts (Zotero Abstracts Workflow)

The following scripts in python_for_Zotero_abstracts are designed for specific types of analysis:

• theory.py: Identify urban planning theory used in abstracts
• n_themes.py: Identify three themes from abstracts
• methods.py: Identify methods used in abstracts
• results.py: Identify results from abstracts
• location.py: Identify where the research was conducted

Each script will prompt you for the required input and provide results in the output Excel file.

How to Use

Custom Analysis Workflow (aiassistedanalysis.py)

1. Prepare Your Data: Place your CSV or Excel file in the data input folder.
2. Run the Script: python ai_assisted_analysis.py
3. Follow Prompts:
   • Select the model to use (or press Enter for the recommended model).
   • Define what you want the program to identify within the text.
   • Choose columns for identifier and content by name.
   • Set the number of runs for analysis.
4. Consensus & Reporting:
   • Optionally aggregate AI responses for consensus and confidence.
   • Optionally append all reporting info to the bottom of the output Excel file.
5. Review Output:
   • Results are saved to an Excel file in the specified output folder.
   • Consensus columns and reporting info are included if selected.

Zotero Abstracts Workflow

The workflow for analyzing Zotero abstracts and bibliographies remains unchanged. Export your collection from Zotero as a CSV or Excel file and use the provided scripts (e.g., methods.py, results.py, location.py, theory.py, n_themes.py) to run targeted analyses on your exported data. Follow the prompts in each script for results and consensus aggregation as before.

Flow Diagram

mermaid graph LR A[Start] --> B[Read CSV/Excel File] B --> C[Initialize Responses List] C --> D{Loop Through Runs} D --> E[Run 1] D --> F[Run 2] D --> G[Run N] E --> H{Loop Through Rows} F --> H G --> H H --> I[Append Title and Abstract] I --> J[Parse prompt] J --> K[Send to AI Model] K --> L[Receive Response] L --> M[Append Response to List] M --> H H --> N[Ensure Correct Number of Elements] N --> O[Create DataFrame] O --> P[Write to Excel File] P --> Q[End]

Directions

1. Navigate to the parent directory: sh cd /path/to/your/directory

2. Create the virtual environment: sh python -m venv venv

3. Activate the virtual environment:

   • On Windows: sh .\venv\Scripts\activate
   • On macOS/Linux: sh source venv/bin/activate

4. Install Requirements: sh pip install -r requirements.txt

5. Install Chosen LLM (gemma2 is the recommended model): sh ollama pull gemma2

Troubleshooting

Import Errors

If you encounter import errors for ollama, pandas, or tqdm, ensure that:

1. Virtual environment is activated: Make sure you've activated your virtual environment before installing packages or running scripts.
2. Correct package installation: Run pip install -r requirements.txt in your activated virtual environment.
3. VS Code Python interpreter: If using VS Code, ensure it's using the Python interpreter from your virtual environment:
   • Open Command Palette (Ctrl+Shift+P)
   • Type "Python: Select Interpreter"
   • Choose the interpreter from your venv folder

Common Package Issues

• ollama package: The correct package name is ollama, not ollama_python
• Excel support: openpyxl is required for writing Excel files with pandas
• Progress bars: tqdm provides the progress bars shown during analysis

Virtual Environment Issues on Windows

If you encounter PowerShell execution policy errors when activating the virtual environment: ```powershell

Use the batch file instead

.\venv\Scripts\activate.bat

Or bypass execution policy temporarily

powershell -ExecutionPolicy Bypass -File .\venv\Scripts\Activate.ps1 ```

Cloning the Repository

1. Open a terminal or command prompt.
2. Navigate to the directory where you want to clone the repository: sh cd /path/to/your/directory
3. Clone the repository: sh git clone https://github.com/hleve/AI_Analysis_Tool.git
4. Navigate to the cloned repository directory: sh cd AI_Analysis_Tool

Running a Python File

1. Ensure the virtual environment is activated:

   • On Windows: sh .\venv\Scripts\activate
   • On macOS/Linux: sh source venv/bin/activate

2. Run the desired Python file: sh python <filename>.py Replace <filename> with the name of the Python file you want to run. For example, to run ai_assisted_analysis.py, use: sh python ai_assisted_analysis.py

3. Follow any additional prompts or instructions provided by the script to complete the analysis.