Transfer Learning in Biomedical Natural Language Processing: An Evaluation of BERT and ELMo on Ten Benchmarking Datasets

Benchmark Model Rank Results
document-classification-on-hocNCBI_BERT(large) (P)#2F1: 87.3
medical-named-entity-recognition-on-shareNCBI_BERT(base) (P+M)#2F1: 0.792
named-entity-recognition-on-bc5cdr-chemicalNCBI_BERT(base) (P)#10F1: 93.5
named-entity-recognition-on-bc5cdr-diseaseNCBI_BERT(base) (P)#4F1: 86.6
natural-language-inference-on-mednliNCBI_BERT(base) (P+M)#5Accuracy: 84.00
relation-extraction-on-chemprotNCBI_BERT(large) (P)#9F1: 74.4
semantic-similarity-on-biossesNCBI_BERT(base) (P+M)#3Pearson Correlation: 0.9159999999999999