Extraction of Definitional Contexts through Machine Learning

Automatic extraction of definitional contexts has been a problem that deserved to be addressed to in different studies by applications demands in the Natural Language Processing. The first approach to the automatic extraction of these resources has been through specific linguistic patterns, but this approach requires previous extensive linguistic knowledge and a thorough previous work. A model machine learning, on the other hand, reduces the work and, as we believe, can improve the results obtained with only one approach based on linguistic rules. Here experiments for extraction/classification of definitional contexts with naive bayes classifier and SVM are presented. We show that through machine learning approaches we can improve the results of this specific task. The highest result was obtained by the naive bayes classifier with back-off as smoothing.