Evaluation Dataset: Samples/words in Bio-SimVerb (verbs) and Bio-SimLex (nouns) are collected from a pre-processed PubMed Central Open Access subset (PMC). POS tags and tokens in this resource are generated using the BLLIP constituency parser, trained on a biomedical corpus. The resource covers over 1.4M full articles with more than 388M parsed sentences.
Details can be referred in Section 3 of the paper: Bio-SimVerb and Bio-SimLex: wide-coverage evaluation sets of word similarity in biomedicine
Software / Usage instructions
Medical Research Council (MR/M013049/1)