Question Classification in the Cancer Domain

Abstract

We are investigating question classification for restricted domains with the broader goal of supporting mixed-initiative interaction on mobile phones. In this thesis, we present the development of a new domain-specific corpus of cancer-related questions, a new taxonomy of Expected Answer types, and our efforts toward training a classifier. This work is the first of its kind in the cancer domain using a corpus consisting of real user questions gathered from cQA websites, and a taxonomy built from that corpus. Our goal is to create software to engage newly diagnosed prostate cancer patients in question-answering dialogs related to their treatment options. We are focusing our work on the interaction environment afforded by text and multimedia (SMS and MMS) messaging using mobile telephones, because of the prevalence of this technology and the growing popularity of text messaging, especially among underserved populations.