Automatic Research Question Generation
Purpose:
The project Automatic Research Question Generation (ARQG) is an initial step towards dealing with the fundamental question of Automatic Idea Generation (AIG): What are the most fruitful questions that might be asked before a novel idea can be formed?
Automatic Question Generation (AQG) has been studied before but typically deals with simple questions with single and short factual answers. ARQG is substantially different in that answers to research questions might not yet exist, can be long, and are often indefinite, requiring knowledge from various domains. If we can ask the right question, it might be the one that, in attempting to answer it, leads to the fertile idea. If we can derive useful research questions from a data set, then this might provide foundational material upon which to work toward the goal of AIG across disparate knowledge domains. A training set from a single knowledge domain comprising abstracts and human-identified research questions and keywords/phrases supporting the questions will be used to train an artificial neural network, following which the network will be assessed on its generation of research questions from unseen abstracts.