National audienceIdentification of units of measures in scientific texts. The work presented in this paper consists in identifying specialized terms (units of measures) in textual documents in order to enrich a onto-terminological resource (OTR). The first step permits to predict the localization of unit of measure variants in the documents. We have used a method based on supervised learning. This method permits to reduce significantly the variant search space staying in an optimal search context (reduction of 86% of the search space on the studied set of documents). The second step uses a new similarity measure identifying automatically variants associated with term denoting a unit of measure already present in the OTR with a precision rat...