A Study on Evaluation Method of Wikipedia's Article Using Network Feature Quantity in Definition Statement Extension for Word Sense Disamibiguation

Abstract

Word Sense Disambiguation is one of the basic tasks of Natural Language Processing. This is the task which distinguishes multi-sense word in a sentence. Wikipedia was regarded as a large scale-network that articles are nodes and links are edges study using the link structure of Wikipedia. Therefore the present study extended define statement of word sense by selecting articles which have strong connectivity with other word sense. Experiment of Words Sense Disambiguation failed in selecting articles and there are word sense whose correct answer rate was low. Therefore the correct answer rate is considered to rise evaluating usefulness of articles selected quantitatively. The present study focused on network feature quantity and analysed the network structure. It was found from the result of the observation that especially value of Modularity $Q$ affected the correct answer rate.