Super-sense tagging using support vector machines and distributional features

Abstract

This paper describes our participation in the EVALITA 2011 Super-Sense Tagging (SST) task. SST is the task of annotating each word in a text with a super-sense that defines a general concept such as animal, person or food. Due to the smaller set of concepts involved the task is simpler than Word Sense Disambiguation one which identifies a specific meaning for each word. In this task, we exploit a supervised learning method based on Support Vector Machines. However, supervised approaches are subject to the data-sparseness problem. This side effect is more evident when lexical features are involved, because test data can contain words with low frequency (or absent) in training data. To overcome the sparseness problem, in our supervised strategy, we incorporate information coming from a distributional space of words built on a large corpus, Wikipedia. The results obtained in the task show the effectiveness of our approach. © Springer-Verlag Berlin Heidelberg 2013.