Answer triggering of factoid questions: A cognitive approach
Kingsley Nketia Acheampong, Zhen-Hao Pan, Erqiang Zhou, Xiaoyu Li · 2016
Answer triggering task as new subtle challenge to Question Answering (QA), requires QA systems to have the ability to detect whether there exists at least one valid answer in the set of candidate sentences for the question; and if yes, select one of the valid answer sentences from the candidate sentence set. This paper presents a novel approach that addresses answer triggering task in answer sentence selection modules. It focuses on selecting valid answers by using a cognitive approach that emulates how people answer multiple choice questions, clarifying whether a candidate sentence is valid to the question being answered or not. Moreover, performance is further improved by buttressing its cognitive techniques using grammatical syntactic structure of lexical categories of a candidate sentence's words and application of deep learning computations of lexical semantic models.