Comment by eamag

Author here, thanks for all feedback! Answering some questions:

- I wanted to publish it asap and get feedback, that's why it's a static website with some ui bugs, you can see a json with comments in a static folder so updates have to be with something like GitHub actions or I have to connect a db

- I included replies on purpose because I saw people saying "I'm also working on this, here's my project" and I wanted to index them

- tags are created with an llm and then there is another deduplication pass to combine them. You can imagine some pros and cons of this approach. I also didn't use embeddings with clustering because it didn't really work for another project of mine https://eamag.me/2024/Automated-Paper-Classification#icml-em...

- it flew under radar first time I posted, but shot to a front page after a second chance pool. If you want to contact me, collaborate, delete data or help to improve this website - there are contact options on my blog and an anonymous feedback form in the footer!