Simple interface for advanced semantic search of papers in the ACL circuit using the ColBERT retriever on paper abstracts. Up to date through January, 2025. Papers and abstracts are taken directly from the ACL master bib file, and we will try to update the index to contain the newest papers as often as we can. Note that (1) GPUs are not required, and (2) papers are indexed based on their abstract, so the tool is not for searching authors/titles. Rather, it's for finding the most relevant papers to any claim, research question, topic, direction, etc.
- Clone the repo, and then use
pipto install the libraries listed inrequirements.txt:
pip install -r requirements.txt- Download the latest data:
chmod +x download_data.sh
./download_data.sh- Run the application on the host server (make sure your host exposes the 5000 port, or change it with the
--portflag):
python app.py --web- Using a browser, connect to
http://YOUR.HOST.IP.ADDRESS:5000- Use the interface to search!
If your host does not allow you to expose a port, then you can run the console version:
python app.py --consoleFeel free to open an issue or submit a pull request for improvements!