/daily_paper_recommendation

Designed to provide me with my daily dose of what ML papers / topics to read

Primary LanguagePythonMIT LicenseMIT

Daily Paper Recommendation

I normally sort arxiv by new and select papers of interest to read every day. I'm hoping to incorporate more ML / DL into this process.

How to Run

  1. Install the pipenv
  2. Launch via uvicorn main:app
  3. View the code for how to make a call to (local api) in sample_response.py

Pushing changes to heroku

  1. heroku create - first time only
  2. git push heroku HEAD:master to push changes + build app

Currently, the process involves:

  1. Hitting the ARXIV API for the most recent papers (will be designed to run each day)
  2. Embed a selected historical list of paper titles
  3. Compute cosine similarity to embeddings of paper titles from today
  4. Return the most similar papers

Some further to dos:

  1. Deploy as an app to be called
  2. Change into a knowledge graph
  3. Evaluate feasibility of considering abstract as well
  4. Automatically hit Mendeley API to retrieve titles list (for some reason, can't get access token ATM)
  5. Return links to papers