基于场景与对话结构的摘要生成研究

doi:10.19678/j.issn.1000-3428.0064382

摘要/Abstract

摘要： 对话摘要是从复杂的对话中提取关键信息以转化成简短的文本，供用户快速浏览对话内容。相比传统文本摘要，对话摘要数据具有篇幅较长、结构复杂等特点。传统的摘要模型难以充分地利用数据的长文本信息，并且无法考虑对话的结构信息。为此，结合抽取和生成模型，提出一种基于场景与对话结构的摘要生成方法，以对话中的场景、角色和对话内容为主要信息生成对话摘要。通过对话解析构建以角色、动作说明和会话为要素的对话结构图，并使用序列标注任务微调BERT预训练模型，生成以对话句子为单位的向量表示，利用图神经网络建模对话结构，筛选出包含关键信息的句子。在此基础上，将得到的抽取结果作为生成模型的输入，利用双向自回归变压器（BART）预训练模型作为基础框架，在编码端额外引入角色和场景信息，丰富生成模型的语义特征，使用加入多头注意力机制的解码器生成摘要。实验结果表明，相比BART、MV_BART、HMNet等方法，该方法在ROUGE-1指标上最高可提升5.3个百分点。

关键词: 对话摘要, 长文本摘要, 文本结构, 对话结构, 双向自回归变压器预训练

Abstract: The dialogue summarization's goal is to condense key information from complex dialogues into concise text, allowing users to browse through the content quickly.Unlike traditional text summarization, dialogue summarization data is characteristically lengthy and complex in structure.The traditional summarization model is unable to exploit the lengthy text information fully, and neglects the structural information of the conversation.To this end, this paper proposes a summarization generation method based on scene and dialogue structure by combining the extraction and generation model.This method generates the dialogue summarization using the scene, role, and dialogue content.Through dialogue analysis, a dialogue structure diagram incorporating elements such as roles, action descriptions, and conversations is constructed.The BERT pre-training model is fine-tuned using sequence tagging tasks to generate vector representation with dialogue sentences as the unit.Graph neural networks model the dialogue structure and filter out sentences containing key information.On this basis, the generation model uses the extracted results as the input.A Bidirectional and Auto-Regressive Transformer(BART) pre-training model as the basic framework, introduces additional role and scene information at the coding end, enriches the semantic features of the generation model, and uses the decoder with a multi-head attention mechanism to generate the summarization.The experimental results indicate that the proposed method can increase the ROUGE-1 index by up to 5.3 percentage points, compared to BART, MV_BART, HMNet, and other methods.

Key words: dialogue summarization, long text summarization, text structure, dialogue structure, Bidirectional and Auto-Regressive Transformers(BART) pre-training

中图分类号:

TP391

李健智, 王红玲, 王中卿. 基于场景与对话结构的摘要生成研究[J]. 计算机工程, 2023, 49(4): 303-311.

LI Jianzhi, WANG Hongling, WANG Zhongqing. Research on Summarization Generation Based on Scene and Dialogue Structure[J]. Computer Engineering, 2023, 49(4): 303-311.

https://www.ecice06.com/CN/Y2023/V49/I4/303

图/表 11

20230417190727

20230417190730

20230417190734

20230417190738

20230417190741

20230417190744

20230417190748

20230417190751

20230417190754

20230417190757

20230417190800

参考文献

[1] 刘挺, 吴岩, 王开铸.自动文摘综述[J].情报科学, 1998, 16(1):63-69. LIU T, WU Y, WANG K Z.Summary of automation abstract[J].Information Science, 1998, 16(1):63-69.(in Chinese)
[2] ROMAN N T, PIWEK P, CARVALHO A M B R.Politeness and bias in dialogue summarization:two exploratory studies[M]//STEFANOV V, TRAIT J.The information retrieval series, Berlin, Germany:Springer, 2006:171-185.
[3] 朱永清, 赵鹏, 赵菲菲, 等.基于深度学习的生成式文本摘要技术综述[J].计算机工程, 2021, 47(11):11-21, 28. ZHU Y Q, ZHAO P, ZHAO F F, et al.Survey on abstractive text summarization technologies based on deep learning[J].Computer Engineering, 2021, 47(11):11-21, 28.(in Chinese)
[4] CHO K, VAN MERRIENBOER B, GULCEHRE C, et al.Learning phrase representations using RNN encoder-decoder for statistical machine translation[C]//Proceedings of Conference on Empirical Methods in Natural Language Processing.Stroudsburg, USA:Association for Computational Linguistics, 2014:1-10.
[5] 王红玲, 周国栋, 朱巧明.面向冗余度控制的中文多文档自动文摘[J].中文信息学报, 2012, 26(2):92-96. WANG H L, ZHOU G D, ZHU Q M.Chinese multi-document summarization based on redundancy control[J].Journal of Chinese Information Processing, 2012, 26(2):92-96.(in Chinese)
[6] PAGE L, BRIN S, MOTWANI R, et al.The PageRank citation ranking:bringing order to the web[C]//Proceedings of Technical Report, Stanford Digital Library Technologies.Stroudsburg, USA:Association for Computational Linguistics, 1998:1-10.
[7] MIHALCEA R, TARAU P.TextRank:bringing order into texts[EB/OL].[2022-03-01].https://www.researchgate.net/profile/Paul-Tarau/publication/200042361_TextRank_Bringing_Order_into_Text/links/0912f508a98af2fe24000000/TextRank-Bringing-Order-into-Text.pdf.
[8] CHENG J P, LAPATA M.Neural summarization by extracting sentences and words[EB/OL].[2022-03-01].https://arxiv.org/pdf/1603.07252.pdf.
[9] LIU Y.Fine-tune BERT for extractive summarization[EB/OL].[2022-03-01].https://arxiv.org/pdf/1903.10318.pdf.
[10] RUSH A M, CHOPRA S, WESTON J.A neural attention model for abstractive sentence summarization[C]//Proceedings of Conference on Empirical Methods in Natural Language Processing.Stroudsburg, USA:Association for Computational Linguistics, 2015:379-389.
[11] GU J T, LU Z D, LI H, et al.Incorporating copying mechanism in sequence-to-sequence learning[C]//Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics.Stroudsburg, USA:Association for Computational Linguistics, 2016:1631-1640.
[12] SEE A, LIU P J, MANNING C D.Get to the point:summarization with pointer-generator networks[C]//Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics.Stroudsburg, USA:Association for Computational Linguistics, 2017:1073-1083.
[13] GULCEHRE C, AHN S, NALLAPATI R, et al.Pointing the unknown words[C]//Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics.Stroudsburg, USA:Association for Computational Linguistics, 2016:140-149.
[14] ZENG W Y, LUO W J, FIDLER S, et al.Efficient summarization with read-again and copy mechanism[EB/OL].[2022-03-01].https://arxiv.org/pdf/1611. 03382.pdf.
[15] COHAN A, DERNONCOURT F, KIM D S, et al.A discourse-aware attention model for abstractive summarization of long documents[EB/OL].[2022-03-01].https://arxiv.org/pdf/1804.05685.pdf.
[16] DEVLIN J, CHANG M W, LEE K, et al.BERT:pre-training of deep bidirectional Transformers for language understanding[EB/OL].[2022-03-01].https://arxiv.org/pdf/1810.04805.pdf.
[17] VASWANI A, SHAZEER N, PARMAR N, et al.Attention is all you need[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems.New York, USA:ACM Press, 2017:5998-6008.
[18] LIU Y, LAPATA M.Text summarization with pretrained encoders[EB/OL].[2022-03-01].https://arxiv.org/abs/1908. 08345.
[19] DONG L, YANG N, WANG W H, et al.Unified language model pre-training for natural language understanding and generation[EB/OL].[2022-03-01].https://arxiv.org/abs/1905.03197.
[20] LEWIS M, LIU Y H, GOYAL N, et al.BART:denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension[C]//Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics.Stroudsburg, USA:Association for Computational Linguistics, 2020:1-10.
[21] 吴仁守, 张宜飞, 王红玲, 等.基于层次结构的生成式自动文摘[J].中文信息学报, 2019, 33(10):90-98. WU R S, ZHANG Y F, WANG H L, et al.Abstractive summarization based on hierarchical structure[J].Journal of Chinese Information Processing, 2019, 33(10):90-98.(in Chinese)
[22] 张迎, 王中卿, 王红玲.基于篇章主次关系的单文档抽取式摘要方法研究[J].中文信息学报, 2019, 33(8):67-76. ZHANG Y, WANG Z Q, WANG H L.Single document extractive summarization with satellite and nuclear relations[J].Journal of Chinese Information Processing, 2019, 33(8):67-76.(in Chinese)
[23] WEI W J, WANG H L, WANG Z Q.Abstractive summarization via discourse relation and graph convolutional networks[C]//Proceedings of International Conference on Natural Language Processing and Chinese Computing.Berlin, Germany:Springer, 2020:331-342.
[24] GOO C W, CHEN Y N.Abstractive dialogue summarization with sentence-gated modeling optimized by dialogue acts[C]//Proceedings of Spoken Language Technology Workshop.Washington D.C., USA:IEEE Press, 2019:735-742.
[25] LI M L, ZHANG L Y, JI H, et al.Keep meeting summaries on topic:abstractive multi-modal meeting summarization[C]//Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics.Stroudsburg, USA:Association for Computational Linguistics, 2019:1-10.
[26] ZHU C G, XU R C, ZENG M, et al.A hierarchical network for abstractive meeting summarization with cross-domain pretraining[C]//Proceedings of Findings of the Association for Computational Linguistics:EMNLP 2020.Stroudsburg, USA:Association for Computational Linguistics, 2020:1-10.
[27] ZHAO L L, XU W R, GUO J.Improving abstractive dialogue summarization with graph structures and topic words[C]//Proceedings of the 28th International Conference on Computational Linguistics.Stroudsburg, USA:International Committee on Computational Linguistics, 2020:1-10.
[28] CHEN J A, YANG D Y.Multi-view sequence-to-sequence models with conversational structure for abstractive dialogue summarization[C]//Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing.Stroudsburg, USA:Association for Computational Linguistics, 2020:1-10.
[29] FENG X C, FENG X C, QIN B, et al.Dialogue discourse-aware graph model and data augmentation for meeting summarization[C]//Proceedings of the 13th International Conference on Artificial Intelligence.Berlin, Germany:[s.n.]:1-15.
[30] FENG X C, FENG X C, QIN B.Incorporating commonsense knowledge into abstractive dialogue summarization via heterogeneous graph networks[C]//Proceedings of China National Conference on Chinese Computational Linguistics.Berlin, Germany:Springer, 2021:127-142.
[31] FENG X C, FENG X C, QIN L B, et al.Language model as an annotator:exploring DialoGPT for dialogue summarization[EB/OL].[2022-03-01].https://arxiv.org/abs/2105.12544.
[32] WU L F, CHEN Y, SHEN K, et al.Graph neural networks for natural language processing:a survey[EB/OL].[2022-03-01].https://arxiv.org/abs/2106.06090.
[33] LIN C Y, HOVY E.Automatic evaluation of summaries using N-gram co-occurrence statistics[C]//Proceedings of Conference of the North American Chapter of the Association for Computational Linguistics on Human Language Technology.Morristown, USA:Association for Computational Linguistics, 2003:150-157.

选择文件类型/文献管理软件名称

选择包含的内容