Semantic role labeling using ensemble classifier

Semantic role labeling is an eminent task in natural language processing consisting of the detection of the semantic arguments associated with the predicate or verb of a sentence and grouping into their specific roles. Effective semantic role labeling (SRL) needs lexical, syntactic and word sense informations. The word-sense disambiguation is an open issue in natural language processing. This work basically concentrates on enhancing the task of SRL by solving word sense disambiguation issue and utilizing ensemble classifier. The word sense disambiguation issue can be kept away by adding the collocation features and selectional preferences to the existing lexical features. The ensemble classifier considers two classifiers such as decision tree and support vector machine classifier. The number of component classifiers of an ensemble has a major role in the accuracy of prediction. The semantic role labeling was done using predicted probabilities of ensemble classifier.