View article

1 Abstract uestion-Classification is the task of identifying the required answer type for any question posed to a uestion-Answering system. The space of required answer types often varies based on the purposes of the A system. So, it is more useful to do a two-level classification of the uestion. We present a machine learning model which achieves this goal using various surface features present in a question. We have tested various features and analyzed the usefulness of each of these for classification purposes. Using a Maxent classifier our system is able to achieve an accuracy of 92% on the coarse-level classes and 86% on finer-level classes.