BioSentVec: creating sentence embeddings for biomedical texts

TL;DR

BioSentVec is a novel set of sentence embeddings trained on over 30 million biomedical documents, improving semantic understanding in biomedical NLP tasks and outperforming existing models.

Contribution

It introduces the first open biomedical sentence embeddings trained on large-scale data, filling a gap in domain-specific NLP resources.

Findings

BioSentVec outperforms other models in sentence similarity tasks.

It achieves state-of-the-art performance in biomedical sentence embeddings.

The embeddings are publicly available for research use.

Abstract

Sentence embeddings have become an essential part of today's natural language processing (NLP) systems, especially together advanced deep learning methods. Although pre-trained sentence encoders are available in the general domain, none exists for biomedical texts to date. In this work, we introduce BioSentVec: the first open set of sentence embeddings trained with over 30 million documents from both scholarly articles in PubMed and clinical notes in the MIMIC-III Clinical Database. We evaluate BioSentVec embeddings in two sentence pair similarity tasks in different text genres. Our benchmarking results demonstrate that the BioSentVec embeddings can better capture sentence semantics compared to the other competitive alternatives and achieve state-of-the-art performance in both tasks. We expect BioSentVec to facilitate the research and development in biomedical text mining and to complement the existing resources in biomedical word embeddings. BioSentVec is publicly available at https://github.com/ncbi-nlp/BioSentVec