Composing Answer from Multi-spans for Reading Comprehension
Lists
Only you see this
Abstract
This paper presents a novel method to generate answers for non-extraction machine reading comprehension (MRC) tasks whose answers cannot be simply extracted as one span from the given passages. Using a pointer network-style extractive decoder for such type of MRC may result in unsatisfactory performance when the ground-truth answers are given by human annotators or highly re-paraphrased from parts of the passages. On the other hand, using generative decoder cannot well guarantee the resulted answers with well-formed syntax and semantics when encountering long sentences. Therefore, to alleviate the obvious drawbacks of both sides, we propose an answer making-up method from extracted multi-spans that are learned by our model as highly confident $n$-gram candidates in the given passage. That is, the returned answers are composed of discontinuous multi-spans but not just one consecutive span in the given passages anymore. The proposed method is simple but effective: empirical experiments on MS MARCO show that the proposed method has a better performance on accurately generating long answers, and substantially outperforms two competitive typical one-span and Seq2Seq baseline decoders.
BibTeX
@misc{zhang2020composing,
author = {Zhuosheng Zhang and Yiqing Zhang and Hai Zhao and Xi Zhou and Xiang Zhou},
title = {{Composing Answer from Multi-spans for Reading Comprehension}},
year = {2020},
eprint = {2009.06141},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2009.06141},
}Loading related content...