search

allenai / s2-folks

Public space for the user community of Semantic Scholar APIs to share scripts, report issues, and make suggestions.
Other
190 stars 29 forks source link

Q: Are paper-embedding-public-apis part of the Semantic Scholar APIs? #132

Closed de-code closed 1 year ago

de-code commented 1 year ago

I am interested in using paper-embedding-public-apis. But I am not quite sure what the expected availability is. Should I also pass in the API key to that API?

cfiorelli commented 1 year ago

@de-code Hello and thank you for the question. It looks like you're seeking to use Specter with S2. We just published news for using Specter in our official release notes as follows:

Specter2 embeddings

AI2’s SPECTER 2.0 is the latest version of our document embedding models designed specifically for scientific tasks. Building upon its predecessor, SPECTER is tailored to generate embeddings for specific tasks, such as Classification, Regression, and Proximity. The API serves pre-computed Proximity embeddings, appropriate for computing the similarity between two documents. The model is hosted as allenai/specter2 on HuggingFace, along with the models for other task-specific embeddings.

Could you review and confirm this resource aligns with what you're looking for?

de-code commented 1 year ago

Thank you for getting back to me. I understand I could use the model via HuggingFace. My original question was more around the API offered to get embedding for title and abstract. e.g. for papers that aren't currently available via the S2 Graph API. Is that API properly supported by S2 or is it just meant for research? And now that you switched to offering SPECTER 2, will that API get updated to support that too?

cfiorelli commented 1 year ago

Looks like the embedding page is not under s2 nor scholar and is probably not supported by S2. As a result of S2 not being the managers of that repo I can't say when or if it would receive the v2 model.

I did notice at the bottom there is a comment about submitting questions directly on that repo - if you end up trying that let me know if that helps. best