Proposal to add the MedSTS dataset

bigscience-workshop / biomedical

Tools for curating biomedical training data for large-scale language modeling

458 stars 116 forks source link

Open FremyCompany opened 2 years ago

FremyCompany commented 2 years ago

Name: MedSTS
Description: 1,068 sentence pairs annotated by two medical experts with semantic similarity scores of 0-5 (low to high similarity).
Task: STS
Paper: https://arxiv.org/abs/1808.09397
Data: (must be asked by email to Mayo Clinic)
License: (probably depends on agreement with Mayo Clinic)
Motivation: one of the largest clinical text similarity dataset

hakunanatasha commented 2 years ago

@FremyCompany I recognized you proposed this, will you consider implementing it?