microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract

Pretraining large neural language models has led to impressive gains on many natural language processing (NLP) tasks . Most pretraining efforts focus on general domain corpora, such as newswire and Web . Recent work shows pretraining language models from scratch results in substantial gains over continual pretraining of general-domain language models . This model achieves state-of-the-art performance on several biomedical NLP tasks, as shown on the Biomedical Language Understanding and Reasoning Benchmark . If you find this paper useful in your research, please cite it as a reference to the study published on the ArXiv: 2007.15779. For more information, please contact the authors of this paper,