Abstract
Prior background knowledge is essential for human reading and understanding. In this work, we investigate how to leverage external knowledge to improve question answering. We primarily focus on multiple-choice question answering tasks that require external knowledge to answer questions. We investigate the effects of utilizing external in-domain multiple-choice question answering datasets and enriching the reference corpus by external out-domain corpora (i.e., Wikipedia articles). Experimental results demonstrate the effectiveness of external knowledge on two challenging multiple-choice question answering tasks: ARC and OpenBookQA.
Abstract (translated by Google)
URL
http://arxiv.org/abs/1902.00993