Strategies d'Identification des Noms Propres a Partir de Nomenclatures Medicales Paralleles.
Bodenreider O, Zweigenbaum P
Traitement Automatique des Langues. 2000;41(3):727-757.
Abstract:
Some specialized domains, such as medicine, make extensive use of proper names in their terminology. The present work studies criteria which can help to classify words as proper names in biomedical terminologies. Useful characteristics include word capitalization, the presence of words in several translations and patterns in terms. Since no individual criterion achieves both high precision and high recall, we propose a combination of five criteria that allowed us to reach 86% precision and recall on one terminology and 98% on another one. Strategic implications are discussed.
Bodenreider O, Zweigenbaum P. Strategies d'Identification des Noms Propres a Partir de Nomenclatures Medicales Paralleles.
Traitement Automatique des Langues. 2000;41(3):727-757.