You are here
MetaMap in the CALBC Workshop II.
MetaMap is a tool which maps biomedical text to UMLS® Metathesauru® concepts. In MetaMap, input text undergoes a lexical/syntactic analysis consisting of a first analysis in which tokens, sentence boundaries and acronyms or abbreviations are identified and each token is assigned a part of speech. Input words are mapped to the SPECIALIST lexicon using lexical lookup and then the SPECIALIST minimal commitment parser identifies phrases and their lexical heads. The identified phrases are processed to generate variants (normally by table lookup), then candidates (Metathesaurus strings) are identified computing and evaluating their match to the input text. Then mapping constructions are produced in which candidates found in the previous step are combined and evaluated to produce a final result that best matches the phrase text. Finally word sense disambiguation (WSD) might optionally be used, in which mappings involving concepts that are semantically consistent with surrounding text are favored. MetaMap is available from http://metamap.nlm.nih.gov/. Downloads are restricted and require a valid UMLSKS user account. A 2011 MEDLINE baseline annotation with MetaMap is available from http://skr.nlm.nih.gov/.