Scientific PDF Analysis Tool

A Python application that processes scientific papers in PDF format and generates detailed scientific analyses using AI models (OpenAI GPT or DeepSeek).

Features

• Text extraction from PDF files
• Scientific analysis using AI models (OpenAI GPT-4/3.5 or DeepSeek)
• Multiple AI provider support
• Batch PDF processing
• Automatic result reporting
• Error handling and retry mechanism
• Configurable model selection

Installation

1. Clone the project:

git clone [repository-url]

2. Install required packages:

pip install -r requirements.txt

3. Create a .env file in the project root directory:
   • Copy .env.example to .env
   • Add your API keys (OpenAI and/or DeepSeek)
   • Select your preferred API provider and model

Example .env file:

# API Configuration
OPENAI_API_KEY=your-openai-api-key-here
DEEPSEEK_API_KEY=your-deepseek-api-key-here

# Model Selection
API_PROVIDER=openai     # Options: openai, deepseek
MODEL_NAME=gpt-4       # OpenAI options: gpt-4, gpt-3.5-turbo
                      # DeepSeek options: deepseek-chat, deepseek-coder

You can get your API keys from:

• OpenAI API key: https://platform.openai.com/api-keys
• DeepSeek API key: https://platform.deepseek.ai/api-keys

Usage

1. Place your PDF files in the Papers directory:

   • The program will process all PDF files in this directory
   • Files are processed sequentially in alphabetical order
   • Each PDF is analyzed independently

2. Run the program:

python gpt_paper.py

3. Check results in the Result directory:
   • For each PDF, a corresponding .txt file is created
   • Output files are named as: [original_pdf_name]_output.txt
   • Results include detailed scientific analysis in a structured format

Important Notes

API Token Usage

• Each PDF analysis consumes API tokens
• Token consumption depends on:
  • Length of the PDF document
  • Selected model (GPT-4 costs more than GPT-3.5-turbo)
  • Selected API provider (pricing varies between OpenAI and DeepSeek)
  • Number of API calls made
• Monitor your API usage at:
  • OpenAI: https://platform.openai.com/usage
  • DeepSeek: https://platform.deepseek.ai/usage

Processing Time

• Processing time varies based on:
  • PDF size and complexity
  • Number of files being processed
  • API response time
  • Rate limiting and retry mechanisms
  • Selected API provider and model

Best Practices

• Start with a small number of PDFs to test the system
• Monitor the console output for processing status
• Keep PDFs in English for best results
• Ensure PDFs are text-searchable (not scanned images)
• Check your API keys have sufficient credits before processing large batches
• Compare results between different models and providers

Output Format

The analysis report generated for each PDF includes:

1. Article Citation

   • Title
   • Authors
   • Journal
   • Volume and Issue
   • Publication Date
   • DOI
   • Publisher

2. Research Purpose and Hypothesis

   • Research Topic
   • Hypothesis/Problem Statement

3. Participants and Study Area

   • Participant Information
   • Study Area

4. Methodology

   • Data Collection Method
   • Tools/Instruments Used
   • Data Analysis Method

5. Results

   • Key Findings
   • Statistical Results

6. Authors' Recommendations and Discussion

   • Research Success Status
   • Authors' Recommendations and Future Research

7. Scientific Contribution and Strengths/Weaknesses

   • Strengths
   • Weaknesses

8. Summary and Scientific Evaluation

Author

Burak Can KARA Email: burakcankara@gmail.com

License

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