arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1808.09384cs.CLcs.AI
收藏

是什么让阅读理解题更难?

What Makes Reading Comprehension Questions Easier?

The University of Tokyo(东京大学);RIKEN AIP(理化学研究所AIP);Tohoku University(东北大学);National Institute of Informatics(国立信息学研究所)

Saku Sugawara, Kentaro Inui, Satoshi Sekine, Akiko Aizawa

摘要:

为机器阅读理解(MRC)创建数据集的一个挑战是收集需要对语言进行复杂理解的问题,而不仅仅是使用表面线索来回答。在这项工作中,我们调查了是什么让最近12个MRC数据集的问题更容易,有三种问题风格(答案提取,描述和多项选择)。我们建议采用简单的算法将每个数据集分为简单和困难的子集,并检查每个子集的两个基线模型的性能。然后,我们手动注释从每个子集的有效性和必要的推理技能,调查哪些技能解释简单和困难的问题之间的差异采样的问题。从这项研究中,我们观察到(i)与整个数据集相比,硬子集的基线表现显着下降,(ii)与简单问题相比,硬问题需要知识推理和多句推理,(iii)多项选择题往往需要比答案提取和描述问题更广泛的推理技能。这些结果表明,人们可能高估了MRC的最新进展。

英文摘要:

A challenge in creating a dataset for machine reading comprehension (MRC) is to collect questions that require a sophisticated understanding of language to answer beyond using superficial cues. In this work, we investigate what makes questions easier across recent 12 MRC datasets with three question styles (answer extraction, description, and multiple choice). We propose to employ simple heuristics to split each dataset into easy and hard subsets and examine the performance of two baseline models for each of the subsets. We then manually annotate questions sampled from each subset with both validity and requisite reasoning skills to investigate which skills explain the difference between easy and hard questions. From this study, we observed that (i) the baseline performances for the hard subsets remarkably degrade compared to those of entire datasets, (ii) hard questions require knowledge inference and multiple-sentence reasoning in comparison with easy questions, and (iii) multiple-choice questions tend to require a broader range of reasoning skills than answer extraction and description questions. These results suggest that one might overestimate recent advances in MRC.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑