🧾 URL Extraction Performance Across arXiv File Formats

This repository contains all data, code, and results related to our study on extracting and evaluating URLs from multi-format representations of arXiv research papers. It supports a longitudinal and format-wise analysis of URL extraction from open-access scholarly documents and includes a pilot dataset of arXiv papers across formats (PDF, LaTeX, HTML, XML, and plain text), ground truth annotations of valid and OADS-related URLs, as well as scripts and the jupyter notebook to extract, evaluate, and visualize URL extraction performance.

---

📂 Repository Structure

├── data/ # Full-text files from arXiv in multiple formats
│  ├── html/ 
│  ├── latex/ 
│  ├── pdf/ 
│  ├── text/ 
│  └── xml/ 
│
├── figures/ 
│
├── results/ 
│  ├── extracted_urls_1000_per_year.json
│  ├── extracted_urls_1000_per_year_10_samples_all_12_folders.json
│  ├── html_urls.json
│  ├── latex_urls.json
│  ├── text_urls.json
│  └── xml_urls.json
│
├── scripts/
│  ├── convert_pdf_using_grobid.py
│  ├── pdf_to_text_converter_arxiv.py
│  └── convert_latex_to_html.sh
│
├── arxiv_extracted_urls_comparison.xlsx 
├── arxiv_file_formats.ipynb
└── README.md

---

📁 Data

• The data/ folder includes the same arXiv papers in:

  • pdf/: original PDFs
  • latex/: LaTeX source files
  • html/: converted using LaTeXML
  • xml/: converted using GROBID
  • text/: plain text via PyMuPDF

• *.json files in results/ contain extracted URL lists by format.

• arxiv_extracted_urls_comparison.xlsx summarizes format coverage and valid URL extractions.

---

⚙️ Key Scripts

scripts/{File}	Description
pdf_to_text_converter_arxiv.py	Converts PDFs to plain text using PyMuPDF
convert_pdf_using_grobid.py	Extracts XML from PDFs using GROBID
convert_latex_to_html.sh	Converts LaTeX source to HTML using LaTeXML

---

🛠️ Tools Used

• Python 3.10.16
• LaTeXML 0.8.8
• GROBID 0.8.1
• PyMuPDF 1.24.13

---

🚀 To reproduce the results

1. Clone the Repository

git clone https://github.com/lamps-lab/arxiv-urls.git
cd arxiv-urls

2. Install Requirements

• Create a virtual environment and install required packages:

  pip install PyMuPDF==1.24.13 lxml pylatexenc
  

• Install LaTeXML 0.8.8 by following the official instructions: https://math.nist.gov/~BMiller/LaTeXML/

3. Run the Jupyter Notebook

• arxiv_file_formats.ipynb – Random paper selection, Format conversion, url extraction, visualizations

---

Rochana R. Obadage 
Updated on: 09/10/2025