Utilization of common human queries in ranking automatically generated questions

Abstract

We challenge a form of Paragraph-to-Question generation task. We propose a question generation system which can be used to generate questions from a body of text. Our goal is to rank the generated questions by using Community-based Question Answering systems to calculate the importance of the questions beside tree kernel functions to assess how grammatically correct they are. The main assumption that our project is based on is that each body of text is related to a topic of interest and it has comprehensive information about the topic.