Simple or Complex? Classifying Questions by Answering Complexity

Simple questions require small snippets of text as the answers whereas complex questions require inferencing and synthesizing information from multiple documents to have multiple sentences as the answers. The traditional QA systems can handle simple questions easily but complex questions often need more sophisticated treatment e.g. question decomposition. Therefore, it is necessary to automatically classify an input question as simple or complex to treat them accordingly. We apply two machine learning techniques and a Latent Semantic Analysis (LSA) based method to automatically classify the questions as simple or complex.