Propositions pour la détection automatique de la néologie sémantique

Various techniques have been developed and implemented to process neology of form automatically, but the processing of semantic neology is still a tricky question. Mechanisms underlying semantic neology are described using semantic features or semes in works on text semantics (Rastier and Valette 2009). In this paper, we propose a prospective approach to detect semantic neology using statistical methods and corpus analysis. These methods are inspired from text semantics, since they use descriptors considered as semes. We propose a process relying on semantic tagging to detect semantic neologisms. This process is applied to study the French adjective toxique in the context of the economic crisis in 2008 and 2009.