Towards a Mixed Approach to Extract Biomedical Terms from Text Corpus

Abstract

The objective of this paper is to present a methodology to extract and rank automatically biomedical terms from free text. The authors present new extraction methods taking into account linguistic patterns specialized for the biomedical domain, statistic term extraction measures such as C-value and statistic keyword extraction measures such as Okapi BM25, and TFIDF. These measures are combined in order to improve the extraction process and the authors investigate which combinations are the more relevant associated to different contexts. Experimental results show that an appropriate harmonic mean of C-value associated to keyword extraction measures offers better precision, both for single-word and multi-words term extraction. Experiments describe the extraction of English and French biomedical terms from a corpus of laboratory tests available online. The results are validated by using UMLS in English and only MeSH in French as reference dictionary.