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.
- Install the pipenv
- Launch via
uvicorn main:app
- View the code for how to make a call to (local api) in
sample_response.py
heroku create
- first time onlygit push heroku HEAD:master
to push changes + build app
Currently, the process involves:
- Hitting the ARXIV API for the most recent papers (will be designed to run each day)
- Embed a selected historical list of paper titles
- Compute cosine similarity to embeddings of paper titles from today
- Return the most similar papers
Some further to dos:
- Deploy as an app to be called
- Change into a knowledge graph
- Evaluate feasibility of considering abstract as well
- Automatically hit Mendeley API to retrieve titles list (for some reason, can't get access token ATM)
- Return links to papers