A Comparative Study of Various Oversampling Techniques for dealing With Tiny Sense Classes in Preposition Sense Disambiguation

Abstract

Abstract: In this paper we compared different oversampling techniques for improving the class wise accuracy of tiny sense classes in Preposition Sense Disambiguation (PSD) in the  English-Malayalam machine translation context. We used preposition ‘with’ as use case and Support Vector Machine (SVM) for classification after oversampling the input data. The analysis showed that the oversampling methods achieved significant improvement in both class-wise accuracy and overall accuracy.