ResearchDigest

Disclaimer: This is a proof-of-concept implementation of a custom arXiv paper fetcher. It is not production-ready and has limitations.

---

Overview

ResearchDigest is a Streamlit-based application that allows users to:

• Fetch recent papers from arXiv (default category: cs.RO - Robotics).
• Download PDFs and extract their text using PyMuPDF (fitz).
• Summarize the extracted text using a Hugging Face t5-small summarization pipeline.
• Categorize papers into basic robotics subdomains (e.g., Control Systems, Robot Vision, Robot Learning).
• Display summaries and categories in an interactive Streamlit dashboard.

---

Screenshots

Below are some screenshots of the application in action:

• Dashboard Overview
  

• Paper Summaries
  

---

Running the Application

To run the application on Windows, use the following command in your terminal:

python -m streamlit run .\main.py

Make sure you have all the required dependencies installed before running the command.

Configuration

• Category Query: By default, the app fetches papers from the cs.RO (Robotics) category. To change this, modify the fetch_arxiv_papers(query="cs.RO", ...) function in main.py.
• Summarization Model: The app uses the t5-small model for summarization. You can switch to a different model by updating the pipeline("summarization", model="...") argument in the load_summarizer() function.
• Categorization Keywords: Edit the categories dictionary in the categorize_paper() function to refine subdomain classification.

---

Limitations

• Error Handling: Minimal error handling is implemented. The app may fail silently on download or extraction errors.
• Summary Lengths: Summaries are truncated to the first 5,000 characters of extracted text.
• Basic Categorization: Keyword-based classification may misclassify papers.
• Model Performance: The t5-small model is lightweight but may produce less accurate summaries for complex texts.

---

Contributing

Contributions are welcome! Feel free to fork the repository and:

• Improve error handling.
• Add more categories or refine keyword matching.
• Integrate more powerful summarization models.

---

License

This project is licensed under the MIT License.