English

Composing Answer from Multi-spans for Reading Comprehension

Computation and Language 2021-08-24 v2 Artificial Intelligence Information Retrieval

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 nn-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.

Keywords

Cite

@article{arxiv.2009.06141,
  title  = {Composing Answer from Multi-spans for Reading Comprehension},
  author = {Zhuosheng Zhang and Yiqing Zhang and Hai Zhao and Xi Zhou and Xiang Zhou},
  journal= {arXiv preprint arXiv:2009.06141},
  year   = {2021}
}

Comments

Due to the policy of our institute, with the agreement of all of the author, we decide to withdraw this paper

R2 v1 2026-06-23T18:30:31.050Z