2.765

2022影响因子

(CJCR)

  • 中文核心
  • EI
  • 中国科技核心
  • Scopus
  • CSCD
  • 英国科学文摘

留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

基于阅读技巧识别和双通道融合的机器阅读理解方法

彭伟 胡玥 李运鹏 谢玉强 牛晨旭

彭伟, 胡玥, 李运鹏, 谢玉强, 牛晨旭. 基于阅读技巧识别和双通道融合的机器阅读理解方法. 自动化学报, xxxx, xx(x): x−xx doi: 10.16383/j.aas.c220983
引用本文: 彭伟, 胡玥, 李运鹏, 谢玉强, 牛晨旭. 基于阅读技巧识别和双通道融合的机器阅读理解方法. 自动化学报, xxxx, xx(x): x−xx doi: 10.16383/j.aas.c220983
Peng Wei, Hu Yue, Li Yun-Peng, Xie Yu-Qiang, Niu Chen-Xu. A machine reading comprehension approach based on reading skill recognition and dual channel fusion machanism. Acta Automatica Sinica, xxxx, xx(x): x−xx doi: 10.16383/j.aas.c220983
Citation: Peng Wei, Hu Yue, Li Yun-Peng, Xie Yu-Qiang, Niu Chen-Xu. A machine reading comprehension approach based on reading skill recognition and dual channel fusion machanism. Acta Automatica Sinica, xxxx, xx(x): x−xx doi: 10.16383/j.aas.c220983

基于阅读技巧识别和双通道融合的机器阅读理解方法

doi: 10.16383/j.aas.c220983
基金项目: 国家自然科学基金(62006222), 国家自然科学基金(U21B2009), 中国科学院战略性先导研究计划(XDC02030400)资助
详细信息
    作者简介:

    彭伟:中关村实验室助理研究员. 2023年获得中国科学院信息工程研究所博士学位.主要研究方向为对话生成, 网络空间安全. E-mail: pengwei@iie.ac.cn

    胡玥:中国科学院信息工程研究所研究员. 主要研究方向为自然语言处理, 人工智能. 本文通信作者. E-mail: huyue@iie.ac.cn

    李运鹏:中国科学院信息工程研究所博士研究生. 2019年获得山东大学计算机科学与技术学院学士学位.主要研究方向为自然语言处理. E-mail: liyunpeng@iie.ac.cn

    谢玉强:中国科学院信息工程研究所博士研究生. 2023年获得中国科学院大学博士学位.主要研究方向为自然语言处理, 认知建模. E-mail: yuqiang.xie@kunlun-inc.com

    牛晨旭:中国科学院信息工程研究所博士研究生. 2021年获得西安电子科技大学计算机科学与技术学院学士学位.主要研究方向为自然语言处理. E-mail: niuchenxu@iie.ac.cn

A Machine Reading Comprehension Approach Based on Reading Skill Recognition and Dual Channel Fusion Machanism

Funds: Supported by National Natural Science Foundation of China (62006222), National Natural Science Foundation of China (U21B2009), and Strategic Priority Research Program of Chinese Academy of Science (XDC02030400)
More Information
    Author Bio:

    PENG Wei Assistant researcher at Zhongguancun Laboratory. He received his Ph.D. degree from University of Chinese Academy of Science in 2023. His research interest covers dialog generation and cyber security

    HU Yue Researcher at the Institute of Information Engineering, Chinese Academy of Sciences. Her research interest covers natural language processing and artificial intelligence. Corresponding author of this paper

    LI Yun-Peng Ph. D. candidate at the Institute of Information Engineering, Chinese Academy of Sciences. He received his bachelor degree from Shandong University in 2019. His research interest covers natural language processing

    XIE Yu-Qiang Ph. D. at the Institute of Information Engineering, Chinese Academy of Sciences. He received his Ph.D. degree from University of Chinese Academy of Science in 2023. His research interest covers natural language generation, cognitive modeling

    NIU Chen-Xu Ph. D. candidate at the Institute of Information Engineering, Chinese Academy of Sciences. She received her bachelor degree from Xidian University in 2021. Her research interest covers natural language processing

  • 摘要: 机器阅读理解任务旨在要求系统对于给定的文章进行理解, 然后针对给定的问题进行回答. 先前的工作重点聚焦在问题和文章之间的交互信息. 然而, 忽略了对问题进行更加细粒度的分析, 如: 问题所考察的阅读技巧是什么. 受到先前阅读理解相关文献的启发, 人类对于问题的理解是一个多维度的过程, 首先人类需要理解问题的上/下文语义信息, 然后再针对不同类型的问题识别其需要使用的阅读技巧, 最后再通过与文章的交互来回答出问题的答案. 出于该目的, 提出一种基于阅读技巧识别和双通道融合的机器阅读理解方法来对问题进行更细致的分析, 从而提高模型回答问题的准确性. 具体来说, 阅读技巧识别器通过对比学习的方法能够显式地捕获阅读技巧的语义信息. 双通道融合机制会将问题与文章的交互信息和阅读技巧的语义信息进行深层次的融合, 从而达到辅助系统理解问题和文章的目的. 为了验证模型的效果, 在FairytaleQA数据集上面进行实验, 实验结果表明, 提出的方法实现了目前在机器阅读理解任务和阅读技巧识别任务上最好的效果.
  • 图  1  在FairytaleQA中的一个例子

    Fig.  1  An example in FairytaleQA

    图  2  模型总体结构图

    Fig.  2  Overall structure of the model

    图  3  阅读技巧识别器的总体结构

    Fig.  3  The structure of the reading skill recognizer

    图  4  双通道融合层结构图

    Fig.  4  Structure of dual channel fuspion layer

    图  5  双通道融合机制的性能比较示意图

    Fig.  5  The performances on the dual channel fusion mechanism

    图  6  三种不同融合机制的比较

    Fig.  6  The comparison on the three different fusion mechanisms

    图  7  技巧识别的可视化示意图

    Fig.  7  The visualization of the skill recognition

    表  1  FairytaleQA数据集的主要统计数据, 该数据集有278本书包含10580个QA对 (个)

    Table  1  Core statistics of the FairytaleQA dataset, which has 278 books and 10580 QA-pairs (numbers)

    FairytaleQA数据集统计指标
    均值标准偏差最小值最大值
    每个故事章节数15.69.8260
    每个故事单词数2305.41480.82287577
    每个章节单词数147.760.012447
    每个故事问题数41.729.15161
    每个章节问题数2.92.380018
    每个问题单词数10.53.2327
    每个答案单词数7.25.8170
    下载: 导出CSV

    表  2  在FairytaleQA中的验证和测试集上的性能 (%)

    Table  2  Performance on the validation and the test set in FairytaleQA (%)

    模型名称FairytaleQA 验证集FairytaleQA 测试集
    B-1B-2B-3B-4ROUGE-LMETEORB-1B-2B-3B-4ROUGE-LMETEOR
    轻量化模型
    Seq2Seq25.126.672.010.8113.616.9426.336.722.170.8114.557.34
    CAQA-LSTM28.058.243.661.5716.158.1130.048.854.171.9817.338.60
    Transformer21.874.941.530.5910.326.0121.725.211.740.6710.276.22
    预训练语言模型
    DistilBERT9.708.20
    BERT10.409.70
    BART19.137.923.422.1412.256.5121.058.933.902.5212.666.70
    微调模型
    BART-Question-types49.10
    CAQA-BART52.5944.1742.7640.0753.2028.3155.7347.0043.6840.4555.1328.80
    BART-NarrativeQA45.3439.1736.3334.1047.3924.6548.1341.5038.2636.9749.1626.93
    BART-FairytaleQA$ \dagger $51.7443.3041.2338.2953.8827.0954.0445.9842.0839.4653.6427.45
    BART-FairytaleQA ‡51.2843.9641.5139.0554.1126.8654.8246.3743.0239.7154.4427.82
    本文模型54.2147.3844.6543.0258.9929.7057.3649.5546.2342.9158.4830.93
    人类表现65.1064.40
    下载: 导出CSV

    表  3  在FairytaleQA的验证集和测试集上对每个组件进行消融研究的结果 (%)

    Table  3  The performance of ablation study on each component in our model on the validation set and the test set of the FairytaleQA (%)

    模型设置FairytaleQA 验证集FairytaleQA 测试集
    B-1B-2B-3B-4ROUGE-LMETEORB-1B-2B-3B-4ROUGE-LMETEOR
    SOTA 模型51.2843.9641.5139.0554.1126.8654.8246.3743.0239.7154.4427.82
    去除阅读技巧识别器52.1544.4742.1140.7355.3827.4554.9047.1643.5540.6756.4829.31
    去除对比学习损失53.2045.0742.8841.9456.7528.1555.2247.9844.1341.4257.3430.20
    去除双通道融合层52.5845.3843.1541.6257.2227.7555.7948.2044.9641.2857.1229.88
    本文模型54.2147.3844.6543.0258.9929.7057.3649.5546.2342.9158.4830.93
    下载: 导出CSV

    表  4  基于交叉熵的方法和基于有监督对比学习的方法在两个任务上的效果 (%)

    Table  4  Performance of the skill recognition and question answering task on cross-entropybased model and contrastive learning (%)

    实验设置ACCB-4R-LMETEOR
    基于交叉熵的方法91.4041.4257.3430.20
    基于有监督对比学习的方法93.7742.9158.4830.93
    下载: 导出CSV

    表  5  不同输入下的阅读技巧识别器的识别表现 (%)

    Table  5  Performance of the skill recognition on different inputs of reading skills recognizer (%)

    实验设置验证集上的准确率测试集上的准确率
    只输入问题85.3182.56
    输入问题和文章92.2493.77
    下载: 导出CSV

    表  6  提出的模型和基线的定性分析示意表

    Table  6  Qualitative analysis of the proposed model and baselines

    文章 国王有个女儿, 和她死去的母亲一样漂亮, 还有一头金黄色的头发. 她长大了 ...... 她将成为王后, 因为国王的女儿和她死去的母亲长得一模一样. 我死后, 她的丈夫将 ......
    问题 国王的女儿长的怎么样?
    预测阅读技巧 人物感知技巧.
    标签 人物感知技巧.
    答案 和她死去的母亲一样漂亮.
    模型预测
      本文模型: 和她死去的母亲一样漂亮, 还有一头金黄色的头发.
      BART-FairytaleQA 和她死去的母亲长得一模一样.
      Transformer: 长得一模一样.
      Seq2Seq: 国王的女儿长得.
    文章 老人和妇人有两头光滑的母牛, 五只母鸡和一只公鸡, 一只老猫和两只小猫. 老人把时间都花在照看奶牛, 母鸡和花园上; 而老妇人则忙着纺纱 ...
    问题 老人把时间花在了什么上?
    预测阅读技巧 动作识别技巧.
    标签 动作识别技巧.
    答案 老人把时间都花在照看奶牛、母鸡和花园上.
    模型预测
      本文模型: 老人把时间都花在照看奶牛、母鸡和花园上.
      BART-FairytaleQA‡: 老人和妇人有两头光滑的母牛, 五只母鸡和一只公鸡.
      Transformer: 照看奶。
      Seq2Seq: 老人把时间都花在看园上.
    下载: 导出CSV
  • [1] Hermann K M, Kociský T, Grefenstette E, Espeholt L, Kay W, Suleyman M, et al. Teaching machines to read and comprehend. In: Proceedings of the Neural Information Processing Systems. Montreal, Canada: 2015. 1693–1701
    [2] Seo M J, Kembhavi A, Farhadi A, Hajishirzi H. Bidirectional attention flow for machine comprehension. arXiv preprint arXiv: 1611.01603, 2016.
    [3] Tay Y, Wang S, Luu A T, Fu J, Phan M C, Yuan X, et al. Simple and effective curriculum pointer-generator networks for reading comprehension over long narratives. In: Proceedings of the Conference of the Association for Computational Linguistics. Florence, Italy: 2019. 4922–4931
    [4] Peng W, Hu Y, Yu J, Xing L X, Xie Y Q. APER: Adaptive evidence-driven reasoning network for machine reading comprehension with unanswerable questions. Knowl. Based Syst., DOI: 10.1016/j.knosys.2021.107364
    [5] Perevalov A, Both A, Diefenbach, D, Ngomo A N. Can machine translation be a reasonable alternative for multilingual question answering systems over knowledge graphs? In: Proceedings of the ACM Web Conference 2022, Lyon, France: 2022. 977–986
    [6] Xu Y, Wang D, Yu M, Ritchie D, Yao B, Wu T, et al. Fantastic questions and where to find them: Fairytaleqa - an authentic dataset for narrative comprehension. In Proceedings of the Conference of the Association for Computational Linguistics. Dublin, Ireland: 2022. 447–460
    [7] Liu S, Zhang X, Zhang S, Wang H, Zhang W. Neural machine reading comprehension: Methods and trends. arXiv preprint arXiv: 1907.01118, 2019.
    [8] Yan M, Xia J, Wu C, Bi B, Zhao Z, Zhang J, et al. A deep cascade model for multi-document reading comprehension. In: Proceedings of the Conference on Artificial Intelligence. Honolulu, USA: 2019. 7354–7361
    [9] Liao J, Zhao X, Li X, Tang J, Ge B. Contrastive heterogeneous graphs learning for multi-hop machine reading comprehension. World Wide Web, 2022, 25(3): 1469–1487 doi: 10.1007/s11280-021-00980-6
    [10] Lehnert W G. Human and computational question answering. Cogn. Sci., 1977, 1(1): 47–73 doi: 10.1207/s15516709cog0101_3
    [11] Kim Y. Why the simple view of reading is not simplistic: Unpacking component skills of reading using a direct and indirect effect model of reading (dier). Scientific Studies of Reading, 2017, 21(4): 310–333 doi: 10.1080/10888438.2017.1291643
    [12] Sugawara S, Yokono H, Aizawa A. Prerequisite skills for reading comprehension: Multi-perspective analysis of mctest datasets and systems. In: Proceedings of the Conference on Artificial Intelligence. San Francisco, USA: 2017. 3089–3096
    [13] Weston J, Bordes A, Chopra S, Mikolov T. Towards ai-complete question answering: A set of prerequisite toy tasks. arXiv preprint arXiv: 1502.05698, 2015.
    [14] Purves A C, Söter A, Takala S and Vähäpassi A. Towards a domain-referenced system for classifying composition assignments. Research in the Teaching of English. 1984. 385–416
    [15] Vähäpassi A. On the specification of the domain of school writing. AFinLAn vuosikirja. 1981. 85–107
    [16] Chen D, Bolton J, Manning C D. A thorough examination of the cnn/daily mail reading comprehension task. arXiv preprint arXiv: 1606.02858, 2016.
    [17] Rajpurkar P, Zhang J, Lopyrev K, Liang P. Squad: 100, 000+ questions for machine comprehension of text. In: Proceedings of the Conference on Empirical Methods in Natural Language Processing. Austin, USA: 2016. 2383–2392
    [18] Kocisk'y T, Schwarz J, Blunsom P, Dyer C, Hermann K M, Melis G. The narrativeqa reading comprehension challenge. Trans. Assoc. Comput. Linguistics, DOI: 10.1162/tacl_a_00023
    [19] Richardson M and Renshaw E. Mctest: A challenge dataset for the open-domain machine comprehension of text. In: Proceedings of the Conference on Empirical Methods in Natural Language Processing. Grand Hyatt Seattle, Washington, USA: 2013. 193–203
    [20] Yang B, Mitchell T M. Leveraging knowledge bases in lstms for improving machine reading. In: Proceedings of the Conference of the Association for Computational Linguistics. Vancouver, Canada: 2017. 1436–1446
    [21] Zhang Z, Wu Y, Zhou J, Duan S, Zhao H, Wang R. Sg-net: Syntax-guided machine reading comprehension. In: Proceedings of the Conference on Artificial Intelligence, New York, USA: 2020. 9636–9643
    [22] Kao K Y, Chang C H. Applying Information Extraction to Storybook Question and Answer Generation. In Proceedings of the Conference on Computational Linguistics and Speech Processing. Taipei, China: 2022. 289–298
    [23] Lu J, Sun X, Li B, Bo L, Zhang T. BEAT: Considering question types for bug question answering via templates. Knowl. Based Syst., DOI: 10.1016/j.knosys.2021.107098
    [24] Wu Z, Xiong Y, Yu S X, Lin D. Unsupervised feature learning via non-parametric instance discrimination. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Salt Lake City, USA: IEEE Computer Society, 2018. 3733–3742
    [25] Chen X, He K. Exploring simple siamese representation learning. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Virtual Event. Computer Vision Foundation / IEEE, 2021. 15750–15758
    [26] Yang J, Duan J, Tran S, Xu Y, Chanda S, et al. Vision-language pre-training with triple contrastive learning. arXiv preprint arXiv: 2202.10401, 2022.
    [27] Chopra S, Hadsell R, LeCun Y. Learning a similarity metric discriminatively, with application to face verification. In: Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition. San Diego, USA: IEEE Computer Society, 2005. 539–546
    [28] Weinberger K Q, Saul L K. Distance metric learning for large margin nearest neighbor classification. J. Mach. Learn. Res., DIO: 10.5555/1577069.1577078 doi: 10.5555/1577069.1577078
    [29] Gao T, et al. Simcse: Simple contrastive learning of sentence embeddings. In: Proceedings of the Conference on Empirical Methods in Natural Language Processing. Virtual Event. 2021. 6894–6910
    [30] Giorgi J M, Nitski O, Wang B, Bader G D. Declutr: Deep contrastive learning for unsupervised textual representations. In: Proceedings of the Conference of the Association for Computational Linguistics. Virtual Event. 2021. 879–895
    [31] Khosla P, Teterwak P, Wang C, Sarna A, Tian Y, Krishnan D, et al. Supervised contrastive learning. In Neural Information Processing Systems. Virtual Event. 2020. 18661–18673
    [32] Li S, Hu X, Lin L, Wen L. Pair-level supervised contrastive learning for natural language inference. In: Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing. Virtual Event. IEEE, 2022. 8237–8241
    [33] Devlin J, Chang M, Lee K, Toutanova K. BERT: pre-training of deep bidirectional transformers for language understanding. In: Proceedings of North American Chapter of the Association for Computational Linguistics. Minneapolis, USA: 2019. 4171–4186
    [34] Papineni K, Roukos S, Ward T, Zhu W. Bleu: a method for automatic evaluation of machine translation. In: Proceedings of the Conference of the Association for Computational Linguistics. Philadelphia, USA: 2002. 311–318
    [35] Lin C Y. Rouge: A package for automatic evaluation of summaries. Text summarization branches out. 2004. 74–81
    [36] Banerjee S, Lavie A. METEOR: an automatic metric for MT evaluation with improved correlation with human judgments. In: Proceedings of the Conference of the Association for Computational Linguistics. Ann Arbor, USA: 2005. 65–72
    [37] Lewis M, Liu Y, Goyal N, Ghazvininejad M, Mohamed A, Levy O, et al. BART: Denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension. In Proceedings of the Association for Computational Linguistics. Virtual Event. 2020. 7871–7880
    [38] Paszke A, Gross S, Chintala S, Chanan G, Yang E, DeVito Z, et al. Automatic differentiation in pytorch. 2017.
    [39] Loshchilov I, Hutter F. Fixing weight decay regularization in adam. arXiv preprint arXiv: 1711.05101. 2017.
    [40] Cho K, Merrienboer B, Bengio Y, Gulcehre C, Bahdanau D, Bougares F, et al. Learning phrase representations using RNN encoder-decoder for statistical machine translation. In Proceedings of the Empirical Methods in Natural Language Processing. Doha, Qatar: 2014. 1724–1734
    [41] Yang C, Jiang M, Jiang B, Zhou W, Li K. Co-Attention Network with Question Type for Visual Question Answering. IEEE Access, DOI: 10.1109/ACCESS.2019.2908035
    [42] Vaswani A, Shazeer N, Parmar N, Uszkoreit J, Jones L, A, et al. Attention is all you need. In: Proceedings of the Neural Information Processing Systems. Long Beach, USA: 2017. 5998–6008
    [43] Sanh V, Debut L, Chaumond, J, Distilbert T W. A distilled version of BERT: smaller, faster, cheaper and lighter. arXiv preprint arXiv: 1910.01108, 2019.
    [44] Mou X, Yang C, Yu M, Yao B, Guo X, Potdar S. Narrative question answering with cutting-edge open-domain QA techniques: A comprehensive study. Trans. Assoc. Comput. Linguistics, DIO: 10.1162/tacl_a_00411 doi: 10.1162/tacl_a_00411
  • 加载中
计量
  • 文章访问数:  90
  • HTML全文浏览量:  40
  • 被引次数: 0
出版历程
  • 网络出版日期:  2023-11-08

目录

    /

    返回文章
    返回