基本篇章单元(elementary discourse units,EDU)识别是构建篇章结构的基础,对篇章分析意义重大。从篇章衔接性视角来看,篇章话题结构理论认为,每个EDU都由要表达信息的起始点(主位)和传达的新信息(述位)两部分构成。因此,EDU识别与主述位识别任务的关系密切。基于此,该文给出了一个基于多任务学习的汉语基本篇章单元和主述位联合识别方法。该方法利用双向长短时记忆网络和图卷积网络对基本单元进行序列化和结构化拓扑信息的表征,再利用多任务学习框架让两个任务共享参数,借助不同任务间的相关性来提升模型的性能。实验结果表明,基于多任务学习的EDU和主述位识别性能均优于单任务学习模型中各自的性能,其中基本篇章单元识别的F1值达到91.90%,主述位识别的F1值达到85.65%。
Abstract
Elementary discourse unit (EDU) recognition is a fundamental task of discourse analysis. From the perspective of discourse cohesion, the theory of discourse topic structure deems that each EDU is closely related to rheme-theme recognition task. Inspired by this notation, this paper proposes a Chinese elementary discourse unit and theme-rheme joint detection method based on multi-task learning. This method applies BiLSTM and Graph Convolutional Networks to represent the EDU’s sequential and structured topological information, and improved the final performance by sharing the parameters of the two model via multi-task learning framework. The experimental results show that the performance of EDU and theme-rheme detection based on multi-task is better than that of the single-task learning model, with the F1-score of up to 91.90% and 85.65%, respectively.
关键词
多任务学习 /
基本篇章单元 /
主位 /
述位
{{custom_keyword}} /
Key words
multi-task /
elementary discourse unit /
theme /
rheme
{{custom_keyword}} /
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] De Beaugrande R, Dressler W U. Introduction to text linguistics[M]. Routledge, 1981.
[2] Carlson L, Marcu D, Okurowski M E. Building a discourse-tagged corpus in the framework of rhetorical structure theory[M].Current and New Directions in Discourse and Dialogue. Springer, Dordrecht, 2003: 85-112.
[3] Prasad Rashmi, Dinesh Nikhil, Lee Alan, et al. The penn discourse treebank 2.0[C]//Proceedings of the International Conference on Language Resources and Evaluation. Marrakech, Morocco, 2008: 2961-2968.
[4] Caroline Sporleder, Mirella Lapata.Discourse chunking and its application to sentence compression[C]//Proceedings of the Conference on Human Language Technology and Empirical Methods in Natural Language Processing.Vancouver, British Columbia, Canada: EMNLP,2005: 257-264.
[5] Ngo Xuan Bach, Nguyen Le Minh, Akira Shimazu.A reranking model for discourse segmentation using subtree features[C]//Proceedings of the 13th Annual Meeting of the Special Interest Group on Discourse and Dialogue (SIGDIAL).Seoul, South Korea: SIGDIAL,2012: 160-168.
[6] Chloe Braud, Ophelie Lacroix, Anders Sgaard.Does syntax help discourse segmentation? Not so much[C]//Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing.Copenhagen, Denmark: EMNLP,2017: 2432-2442.
[7] 李艳翠,冯文贺,周国栋.基于逗号的汉语子句识别研究[J].北京大学学报(自然科学版),2013,29(1): 7-14.
[8] Nianwen Xue, Yaqin Yang. Chinese sentence segmentation as comma classification[C]//Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics.Portland: ACL,2011: 631-635.
[9] Meixun Jin, Mi-Young Kim, Dongil Kim, et al. Segmentation of Chinese long sentences using commas[C]//Proceedings of the 3rd ACL SIGHAN Workshop.Barcelona: ACL,2004: 1-8.
[10] 朱伟华.马泰休斯[J].国外语言学,1987(02): 86-88.
[11] Halliday M A K, Matthiessen C, Halliday M. An introduction to functional grammar[M]. Routledge, 2014.
[12] Park K, Lu X. Automatic analysis of thematic structure in written English[J]. International Journal of Corpus Linguistics, 2015, 20(1): 81-101.
[13] 奚雪峰.汉语篇章话题结构: 表示体系、资源构建及其分析研究[D].苏州: 苏州大学博士学位论文,2017.
[14] Caruana R. Multitask learning[J]. Machine Learning, 1997, 28(1): 41-75.
[15] Zhang Z, Luo P, Loy CC, et al. Facial landmark detection by deep multi-task learning[C]//European Conference on Computer Vision. Springer, Cham, 2014: 94-108.
[16] Liu X, Gao J, He X, et al. Representation learning using multi-task deep neural networks for semantic classification and information retrieval[C]//Proceedings of the NAACL. 2015: 912-921.
[17] Girshick R. Fast r-cnn[C]//Proceedings of the IEEE International Conference on Computer Vision. 2015: 1440-1448.
[18] Arik S ,Chrzanowski M, Coates A, et al. Deep voice: Real-time neural text-to-speech[C]//Proceedings of the 34th International Conference on Machine Learning-Volume 70. JMLR. org, 2017: 195-204.
[19] 徐凡,王明文,谢旭升,等.基于主位-述位结构理论的英文作文连贯性建模研究[J].中文信息学报,2016,30(01): 115-123.
[20] Bastings, J.,Titov, I., Aziz, W, et al. Graph convolutional encoders for syntax-aware neural machine translation.[C]//Proceedings of the EMNLP. 2017: 1957-1967.
[21] Marcheggiani D, Titov I. Encoding sentences with graph convolutional networks for semantic role labeling[J]. arXiv preprint arXiv:1703.04826, 2017.
[22] Linzen T, Dupoux E, Goldberg Y. Assessing the ability of LSTMs to learn syntax-sensitive dependencies[J]. Transactions of the Association for Computational Linguistics, 2016, 4: 521-535.
[23] Li J, Sun A,Joty S. SegBot: A generic neural text segmentation model with pointer network[C]//Proceedings of the IJCAI. 2018: 4166-4172.
[24] Vinyals O, Fortunato M, Jaitly N. Pointer networks[C]//Proceedings of the Advances in Neural Information Processing Systems. 2015: 2692-2700.
[25] Lamb A M,Goyal A G A P, Zhang Y, et al. Professor forcing: A new algorithm for training recurrent networks[C]//Proceedings of Advances in Neural Information Processing Systems. 2016: 4601-4609.
[26] Xuefeng Xi, Guodong Zhou. Building a Chinese discourse topic corpus with micro-topic scheme based on theme-rheme theory[C]//Proceedings of the Big Data Analytics (EI),2017: 1-14.
[27] 李艳翠.汉语篇章结构表示体系及资源构建研究[D].苏州: 苏州大学博士学位论文,2015.
{{custom_fnGroup.title_cn}}
脚注
{{custom_fn.content}}
基金
国家自然科学基金(61876118,61751206)
{{custom_fund}}
PDF(3031 KB)
