Corpus development for Indonesian consumer-health question answering system

Web-based question answering services facilitate users to seek more personalized health-related information. However, the users sometimes have to wait for a while until their questions to be answered. Automatic question answering research can assist the system to generate or retrieve the answer to users. Our work was a pioneering study on consumer-health question answering for Bahasa Indonesia. We built a corpus of 86,731 consumer-health questions, collected from 5 different websites. As part of annotation, we classify the sub-topics for each question, which corresponds to medical specialization. Question sub-topic classification is completed by two complementary approaches: dictionary-based and machine learning-based.