多跳机器阅读理解研究进展

doi:10.19678/j.issn.1000-3428.0061102

摘要/Abstract

摘要： 与常规的单跳机器阅读理解相比，多跳机器阅读理解（MHMRC）需要在多个相关文档段落中进行多跳推理以实现对复杂问题的理解和回答，其更接近于人类的语言推理能力，具有广阔的应用前景但也极具挑战性。阐述MHMRC的研究背景，将现有方法根据适用场景分为封闭集合问答和开放域问答两类，主要包括基于问题分解的方法、基于图神经网络的方法、改进检索的方法、基于推理路径的方法等，分别从模型架构、特点、优劣等方面进行分析。介绍用于多跳推理的非结构化文本数据集和评测指标，对比各模型的性能表现。在此基础上，分析目前MHMRC研究的热点与难点，指出未来发展方向。

关键词: 机器阅读理解, 多跳机器阅读理解, 问题分解, 图神经网络, 开放域问答

Abstract: Compared with common single-hop Machine Reading Comprehension(MRC), Multi-Hop MRC(MHMRC) needs multi-hop reasoning from given multiple documents or paragraphs to understand and answer complex questions.Though MHMRC is extremely challenging, it is closer to human language and reasoning, and has broad application prospects.The research background of MHMRC is introduced and the existing methods are divided according to the applicable scenarios into closed set Question Answering(QA) and Open-domain Question Answering(OpenQA), mainly including methods based on question decomposition, methods based on Graph Neural Network(GNN), methods improving index and methods based on reasoning path, etc.The methods are comprehensively analyzed from the perspectives of model architecture, features, advantages and disadvantages.Then unstructured text datasets and indexes for MHMRC evaluation are described, and they are employed to compare the performance of each model.On this basis, the paper discusses the challenges and hotspots of MHMRC research and the trends of future development are discussed.

Key words: Machine Reading Comprehension(MRC), Multi-Hop MRC (MHMRC), question decomposition, Graph Neural Network(GNN), Open-domain Question Answering(OpenQA)

中图分类号:

TP18

苏珂, 黄瑞阳, 张建朋, 余诗媛, 胡楠. 多跳机器阅读理解研究进展[J]. 计算机工程, 2021, 47(9): 1-17.

SU Ke, HUANG Ruiyang, ZHANG Jianpeng, YU Shiyuan, HU Nan. Research Progress of Multi-Hop Machine Reading Comprehension[J]. Computer Engineering, 2021, 47(9): 1-17.

http://www.ecice06.com/CN/Y2021/V47/I9/1

图/表 18

20210917184842

20210917184846

20210917184850

20210917184853

20210917184857

20210917184900

20210917184904

20210917184907

20210917184911

20210917184915

20210917184920

20210917184923

20210917184927

20210917184932

20210917184936

20210917184940

20210917184944

20210917184948

参考文献

[1] CHEN D Q.Neural reading comprehension and beyond[D].Palo Alto, USA:Stanford University, 2018.
[2] LIU S S, ZHANG X, ZHANG S, et al.Neural machine reading comprehension:methods and trends[EB/OL].(2019-07-02)[2021-02-10].https://arxiv.org/pdf/1907.01118v1.pdf.
[3] MIN S, ZHONG V, ZETTLEMOYER L, et al.Multi-hop reading comprehension through question decomposition and rescoring[C]//Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics.[S.l.]:Association for Computational Linguistics, 2019:6097-6109.
[4] TALMOR A, BERANT J.The Web as a knowledge-base for answering complex questions[C]//Proceedings of 2018 Conference of North American Chapter of the Association for Computational Linguistics:Human Language Technologies.[S.l.]:Association for Computational Linguistics, 2018:1-5.
[5] JIANG Y, BANSAL M.Self-assembling modular networks for interpretable multi-hop reasoning[C]//Proceedings of International Joint Conference on Natural Language Processing.[S.l.]:Association for Computational Linguistics, 2019:4473-4483.
[6] PEREZ E, LEWIS P, CHO K, et al.Unsupervised question decomposition for question answering[C]//Proceedings of 2020 Conference on Empirical Methods in Natural Language Processing.[S.l.]:Association for Computational Linguistics, 2020:8864-8880.
[7] SONG L, WANG Z, YU M, et al.Exploring graph-structured passage representation for multi-hop reading comprehension with graph neural networks[EB/OL].(2018-09-06)[2021-02-10].https://arxiv.org/pdf/1809.02040.pdf.
[8] CAO N, AZIZ W, TITOV I, et al.Question answering by reasoning across documents with graph convolutional networks[C]//Proceedings of 2019 Conference of the North American Chapter of the Association for Computational Linguistics:Human Language Technologies.[S.l.]:Association for Computational Linguistics, 2019:2306-2317.
[9] TU M, WANG G, HUANG J, et al.Multi-hop reading comprehension across multiple documents by reasoning over heterogeneous graphs[C]//Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics.[S.l.]:Association for Computational Linguistics, 2019:2704-2713.
[10] CAO Y, FANG M, TAO D, et al.BAG:bi-directional attention entity graph convolutional network for multi-hop reasoning question answering[C]//Proceedings of 2018 Conference of North American Chapter of the Association for Computational Linguistics:Human Language Technologies.[S.l.]:Association for Computational Linguistics, 2019:357-362.
[11] QIU L, XIAO Y, QU Y, et al.Dynamically fused graph network for multi-hop reasoning[C]//Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics.[S.l.]:Association for Computational Linguistics, 2019:6140-6150.
[12] FANG Y W, SUN S Q, GAN Z, et al.Hierarchical graph network for multi-hop question answering[C]//Proceedings of 2020 Conference on Empirical Methods in Natural Language Processing.[S.l.]:Association for Computational Linguistics, 2020:8823-8838.
[13] THAYAPARAN M, VALENTINO M, SCHLEGEL V, et al.Identifying supporting facts for multi-hop question answering with document graph networks[EB/OL].(2019-10-01)[2021-02-10].https://arxiv.org/pdf/1910.00290.pdf.
[14] TU M, HUANG K, WANG G T, et al.Select, answer and explain:interpretable multi-hop reading comprehension over multiple documents[C]//Proceedings of 2020 AAAI Conference on Artificial Intelligence.[S.l.]:AAAI, 2020:9073-9080.
[15] DING M, ZHOU C, CHEN Q, et al.Exploring graph-structured passage representation for multi-hop reading comprehension with graph neural networks[EB/OL].(2018-09-06)[2021-02-10].https://arxiv.org/pdf/1809.02040.pdf.
[16] YE D, LIN Y, LIU Z, et al.Multi-paragraph reasoning with knowledge-enhanced graph neural network[EB/OL].(2019-11-06)[2021-02-10].https://arxiv.org/pdf/1911.02170.pdf.
[17] TANG Z Y, SHEN Y L, MA X Y, et al.Multi-hop reading comprehension across documents with path-based graph convolutional network[C]//Proceedings of ITCAI-PRICA 2020.Yokohama, Japan:[s.n.], 2020:1-7.
[18] SHAO N, CUI Y M, LIU T, et al.Is graph structure necessary for multi-hop reasoning?[C]//Proceedings of 2020 Conference on Empirical Methods in Natural Language Processing.[S.l.]:Association for Computational Linguistics, 2020:7187-7192.
[19] CHEN D, FISCH A, WESTON J, et al.Reading Wikipedia to answer open-domain questions[C]//Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics.[S.l.]:Association for Computational Linguistics, 2017:1870-1879.
[20] FELDMAN Y, ELYANIV R.Multi-hop paragraph retrieval for open-domain question answering[C]//Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics.[S.l.]:Association for Computational Linguistics, 2019:2296-2309.
[21] MIN S, WALLACE E, SINGH S, et al.Compositional questions do not necessitate multi-hop reasoning[C]//Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics.[S.l.]:Association for Computational Linguistics, 2019:4249-4257.
[22] DAS R, DHULIAWALA S, ZAHEER M, et al.Multi-step retriever-reader interaction for scalable open-domain question answering[C]//Proceedings of International Conference on Learning Representations.New Orleans, USA:DBLP, 2019:1-13.
[23] QI P, LIN X, MEHR L, et al.Answering complex open-domain questions through iterative query generation[C]//Proceedings of International Joint Conference on Natural Language Processing.Berlin, Germany:Springer, 2019:2590-2602.
[24] XIONG W, YU M, GUO X, et al.Simple yet effective bridge reasoning for open-domain multi-hop question answering[C]//Proceedings of EMNLP-IJCNLP 2019.Honk Kong, China:[s.n.]:2019:1-5.
[25] GODBOLE A, KAVARTHAPU D, DAS R, et al.Multi-step entity-centric information retrieval for multi-hop question answering[EB/OL].(2019-09-17)[2021-02-10].https://arxiv.org/pdf/1909.07598.pdf.
[26] NISHIDA K, NAGATA M, MASAAKI N.Answering while summarizing:multi-task learning for multi-hop QA with evidence extraction[C]//Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics.[S.l.]:Association for Computational Linguistics, 2019:2335-2345.
[27] YADAV V, BETHARD S, SURDEANU M, et al.Quick and (not so) dirty:unsupervised selection of justification sentences for multi-hop question answering[EB/OL].(2020-05-03)[2021-02-10].https://arxiv.org/pdf/1911.07176v2.pdf.
[28] QUENTIN G, JULIEN P, ERIC G.Latent question reformulation and information accumulation for multi-hop machine reading[EB/OL].[2021-02-10].https://openreview.net/forum?id=S1x63TEYvr.
[29] CHEN J F, LIN S T, DURRETT G, et al.Multi-hop question answering via reasoning chains[EB/OL].[2021-02-10].https://arxiv.org/pdf/1910.02610.pdf.
[30] JIANG Y, JOSHI N, CHEN Y, et al.Explore, propose, and assemble:an interpretable model for multi-hop reading comprehension[C]//Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics.[S.l.]:Association for Computational Linguistics, 2019:2714-2725.
[31] ASAI A, HASHIMOTO K, HAJISHIRZI H, et al.Learning to retrieve reasoning paths over Wikipedia graph for question answering[EB/OL].(2020-02-14)[2021-02-10].https://arxiv.org/pdf/1911.10470.pdf.
[32] YADAV V, BETHARD S, SURDEANU M.Unsupervised alignment-based iterative evidence retrieval for multi-hop question answering[C]//Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics.[S.l.]:Association for Computational Linguistics, 2020:1-5.
[33] ZHONG V, XIONG C, KESKAR N S, et al.Coarse-grain fine-grain coattention network for multi-evidence question answering[EB/OL].(2019-05-13)[2021-02-10].https://arxiv.org/pdf/1901.00603.pdf.
[34] ZHUANG Y, WANG H.Token-level dynamic self-attention network for multi-passage reading comprehension[C]//Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics.[S.l.]:Association for Computational Linguistics, 2019:1-5.
[35] ZHAO C, XIONG C, ROSSET C, et al.Transformer-XH:multi-evidence reasoning with eXtra hop attention[C]//Proceedings of International Conference on Learning Representations.New Orleans, USA:DBLP, 2019:1-5.
[36] BELTAGY I, PETERS M E, COHAN A.Longformer:the long-document transformer[EB/OL].(2020-12-02)[2021-02-10].https://arxiv.org/pdf/2004.05150v2.pdf.
[37] ZAHEER M, GURUGANESH G, DUBEY A, et al.Big Bird:transformers for longer sequences[EB/OL].(2021-01-08)[2021-02-10].https://arxiv.org/pdf/2007.14062.pdf.
[38] BAUER L, WANG Y, BANSAL M, et al.Commonsense for generative multi-hop question answering tasks[EB/OL].(2019-06-01)[2021-02-10].https://arxiv.org/pdf/1809.06309.pdf.
[39] KOCISKÝ T, SCHWARZ J, BLUNSOM P, et al.The NarrativeQA reading comprehension challenge[EB/OL].(2017-12-19)[2021-02-10].https://arxiv.org/pdf/1712.07040.pdf.
[40] TRIVEDI H, KEON H, KHOT T, et al.Repurposing entailment for multi-hop question answering tasks[C]//Proceedings of 2019 Conference of the North American Chapter of the Association for Computational Linguistics:Human Language Technologies.[S.l.]:Association for Computational Linguistics, 2019:2948-2958.
[41] WELBL J, STENETORP P, RIEDEL S, et al.Constructing datasets for multi-hop reading comprehension across documents[J].Transactions of the Association for Computational Linguistics, 2018, 6:287-302.
[42] JOSHI M, CHOI E, WELD D S, et al.TriviaQA:a large scale distantly supervised challenge dataset for reading comprehension[C]//Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics.[S.l.]:Association for Computational Linguistics, 2017:1601-1611.
[43] YANG Z, QI P, ZHANG S, et al.HotpotQA:a dataset for diverse, explainable multi-hop question answering[C]//Proceedings of 2018 Conference on Empirical Methods in Natural Language Processing.[S.l.]:Association for Computational Linguistics, 2018:2369-2380.
[50] JIANG Y, BANSAL M.Avoiding reasoning shortcuts:adversarial evaluation, training, and model development for multi-hop QA[C]//Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics.[S.l.]:Association for Computational Linguistics, 2019:2726-2736.
[44] JOSHI M, CHEN D, LIU Y, et al.SpanBERT:improving pre-training by representing and predicting spans[J].Transactions of the Association for Computational Linguistics, 2020, 8:64-77.
[45] BACK S, YU S, INDURTHI S R, et al.MemoReader:large-scale reading comprehension through neural memory controller[C]//Proceedings of 2018 Conference on Empirical Methods in Natural Language Processing.[S.l.]:Association for Computational Linguistics, 2018:2131-2140.
[46] CLARK C, GARDNER M.Simple and effective multi-paragraph reading comprehension[C]//Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics.[S.l.]:Association for Computational Linguistics, 2018:1-5.
[47] WEISSENBORN D, KOČISKÝ T, DYER C.Dynamic integration of background knowledge in neural NLU systems[EB/OL].(2018-08-21)[2021-02-10].https://arxiv.org/pdf/1706.02596.pdf.
[48] HU M, PENG Y, HUANG Z, et al.Reinforced mnemonic reader for machine reading comprehension[EB/OL].(2018-06-06)[2021-02-10].https://arxiv.org/pdf/1705.02798v6.pdf.
[49] PAN B, LI H, ZHAO Z, et al.MEMEN:multi-layer embedding with memory networks for machine comprehension[EB/OL].(2017-07-28)[2021-02-10].https://arxiv.org/pdf/1707.09098.pdf.

选择文件类型/文献管理软件名称

选择包含的内容