Descripción y Evaluación de un Sistema de Extracción de Definiciones para el Catalán

Automatic Definition Extraction (DE) consists of identifying definitions in naturally-occurring text. This paper presents a method for the identification of definitions in Catalan in the encyclopedic domain. The train and test corpora come from the Catalan Wikipedia (Viquipèdia). The test set has been manually validated. We approach the task as a supervised classification problem, using the Conditional Random Fields algorithm. In addition to the common linguistic features, we introduce features that exploit the frequency of a word in general and specific domains, in definitional and non-definitional sentences, and in definiendum (term to be defined) and definiens (cluster of words that defines the definiendum) position. We obtain promising results that suggest that combining linguistic and statistical features can prove useful for developing DE systems for under-resourced languages.