Output details

Normalizing biomedical terms by minimizing ambiguity and variability

<24> In biomedicine, high levels of ambiguity and variation in term and entity name forms in text hamper essential mapping to concepts, but efficient, automatic, accurate normalization techniques are hard to achieve. This work is significant in providing a rigorously evaluated novel technique to discover good normalization rules fully automatically, greatly outperforming competitor algorithms. The method is used by e.g. European Bioinformatics Institute for protein name recognition in annotation of protein residues, informs analysis of variation in clinical drug names at National Library of Medicine (US), and is used in services offered by the National Centre for Text Mining.