TOWARDS WORD SENSE DISAMBIGUATION USING MULTIPLE KERNEL SUPPORT VECTOR MACHINE

Word sense disambiguation (WSD), the task of identifying the intended meanings (senses) of words in context, has been a long-standing research objective for natural language processing (NLP). In this paper, we investigate the problem of combining multiple feature channels using kernel methods for the purpose of effective WSD. A straightforward method is to use a uniform combination of more adequate kernels, which are built from different types of data representations or knowledge sources. Instead of using an equal weight for all kernels in the combination, we consider the problem of integrating multiple feature channels using the state-of-the-art multiple kernel learning (MKL) approach, which can learn different weights that reflect the different importance of the feature channels for disambiguation. This approach has the advantage of the possibility to combine and select the more relevant feature channels in an elegant way. Combined with the support vector machine (SVM), this approach is demonstrated with several Senseval/Semeval disambiguation tasks.