Selecting Features for Paraphrasing Question Sentences

In this paper we investigate several schemes for selecting features which are useful for automatically classifying ques tions by their question type We repre sent questions as a set of features and compare the performance of the C machine learning algorithm using the dif ferent representations Experimental re sults show a high accuracy rate in cat egorizing question types using a scheme based on NLP techniques as compared to a scheme based on IR techniques The ultimate goal of this research is to use question type classi cation in order to help identify whether or not two ques tions are paraphrases of each other We hypothesize that the identi cation of fea tures which help identify question type will be useful in the generation of ques tion paraphrases as well