Relink

Inspiration

Broken dataset links make reproducing research difficult. We built ReLink to help researchers reconnect papers with their underlying data

What it does

ReLink discovers papers by topic, extracts research links from PDFs, and investigates missing dataset URLs. It searches Internet Archive and DataCite, ranks candidates, and presents sources, evidence, and verification status. Users can explore datasets associated with discovered papers and export reports.

How we built it

We used Python, Streamlit, httpx, PyMuPDF, and pytest. OpenAlex and arXiv provide papers, while Gemini assists abstract matching and dataset ranking using metadata and small previews.

Challenges we ran into

Broken URLs, API timeouts, rate limits, inconsistent AI responses, and similarly named dataset versions complicated recovery. Distinguishing promising leads from independently verified matches required careful validation.

Accomplishments that we're proud of

We connected paper discovery and dataset investigation in one open-source application.

What we learned

Trustworthy research tools must explain uncertainty, retain original sources, and remain useful when external services fail. Archived webpages alone cannot establish dataset availability.

What's next for Relink

We plan to expand repository coverage, improve version comparison, add durable private-history storage, and strengthen verification through independently sourced checksums.

Built With

• arxiv
• datacite
• gemini
• internet-archive
• openalex
• python
• streamlit