LAUSR

Answer Triggering is still perceived as a challenging task in Question Answering (QA) despite the recent successes chalked up by deep learning models. Its demand for near-human sentence comprehension and answer selection has made previous works on the task seemingly incapable of solving the task. This article introduces an Answer Triggering dataset, CogQA, that contains cognitive features of sentences to enhance the performance of answer triggering systems. It also presents the first deep hierarchical end-to-end neural model that leverages the cognitive elements of CogQA to establish neural correlations to its corresponding answer(s). Our results demonstrate the utility of the dataset and its capability of enabling better triggering of answers in QA systems. Furthermore, our hierarchical model's performance transcends previous works on the WIKIQA benchmark by an appreciable extent.