Empirical methods for the study of denotation in nominalizations

Abstract

This article deals with deverbal nominalizations in Spanish, concretely with the denotative distinction between event and result nominalizations. The goal of this work is twofold: first, to detect the most relevant features for this denotative distinction; and, second, to build an automatic classification system of deverbal nominalizations according to their denotation. We base our study on theoretical hypotheses dealing with this semantic distinction and we analyze them empirically by means of Machine Learning techniques which are at the basis of the ADNClassifier. This is the first tool that aims to automatically classify deverbal nominalizations in event, result or underspecified denotation types in Spanish. The ADN-Classifier has allowed us to empirically evaluate our claims regarding deverbal nominalizations. We set up a series of experiments in order to test the ADN-Classifier with different models and in different realistic scenarios depending on the knowledge resources and natural language processors available. The ADN-Classifier achieves good results (87.20% accuracy).