基于深度学习的自动文本摘要研究综述

doi:10.3778/j.issn.1002-8331.2411-0270

摘要/Abstract

摘要： 自动文本摘要技术是自然语言处理领域的重要研究方向，旨在实现信息的高效压缩与核心语义的保留。随着深度学习技术的快速发展，基于该技术的自动文本摘要方法逐渐成为主流。从抽取式与生成式两大技术路线出发，系统梳理了序列标注、图神经网络、预训练语言模型、序列到序列模型和强化学习等技术在自动文本摘要中的应用，并分析了各类模型的优缺点；介绍了自动文本摘要领域常用的公开数据集、国内低资源语言数据集及评价指标。通过多维度实验对比分析总结了现有技术面临的问题，提出了相应的改进方案。最后，探讨了自动文本摘要的未来研究方向，为后续研究提供参考。

关键词: 自动文本摘要, 深度学习, 生成式摘要, 抽取式摘要, 自然语言处理

Abstract: Automatic text summarization is a significant research direction in the field of natural language processing, aiming to achieve efficient compression of information while preserving its core semantics. With the rapid development of deep learning techniques, methods based on these technologies have gradually become the mainstream. From both extractive and abstractive approaches, this paper systematically reviews the application of techniques such as sequence labeling, graph neural networks, pre-trained language models, sequence-to-sequence models, and reinforcement learning in text summarization, analyzing the strengths and weaknesses of various models. In addition, commonly used public datasets, domestic low-resource language datasets, and evaluation metrics in the field of text summarization are introduced. Multi-dimensional experimental comparisons and analyses are conducted to summarize the current challenges faced by existing technologies and to propose corresponding improvement strategies. Finally, future research directions in automatic text summarization are discussed to provide a reference for subsequent studies.

Key words: automatic text summarization, deep learning, abstractive summarization, extractive summarization, natural language processing

其其日力格, 斯琴图, 王斯日古楞. 基于深度学习的自动文本摘要研究综述[J]. 计算机工程与应用, 2025, 61(18): 24-40.

QI Qirilige, SI Qintu, WANG Siriguleng. Survey of Automatic Text Summarization Based on Deep Learning[J]. Computer Engineering and Applications, 2025, 61(18): 24-40.

参考文献

[1] LUHN H P. The automatic creation of literature abstracts[J]. IBM Journal of Research and Development, 1958, 2(2): 159-165.
[2] 李金鹏, 张闯, 陈小军, 等. 自动文本摘要研究综述[J]. 计算机研究与发展, 2021, 58(1): 1-21.
LI J P, ZHANG C, CHEN X J, et al. Survey on automatic text summarization[J]. Journal of Computer Research and Development, 2021, 58(1): 1-21.
[3] YADAV A K, RANVIJAY, YADAV R S, et al. State-of-the-art approach to extractive text summarization: a comprehensive review[J]. Multimedia Tools and Applications, 2023, 82(19): 29135-29197.
[4] ZHANG M L, ZHOU G, YU W T, et al. A comprehensive survey of abstractive text summarization based on deep learning[J]. Computational Intelligence and Neuroscience, 2022: 7132226.
[5] NALLAPATI R, ZHAI F F, ZHOU B W. SummaRuNNer: a recurrent neural network based sequence model for extractive summarization of documents[C]/Proceedings of the 31st AAAI Conference on Artificial Intelligence, 2017: 3075-3081.
[6] ZHOU Q Y, YANG N, WEI F R, et al. Neural document summarization by jointly learning to score and select sentence [J]. arXiv:1807.02305, 2018.
[7] LIU Y. Fine-tune BERT for extractive summarization[J]. arXiv:1903.10318, 2019.
[8] RUSH A M, CHOPRA S, WESTON J. A neural attention model for abstractive sentence summarization[J]. arXiv:1509. 00685, 2015.
[9] BAHDANAU D, CHO K, BENGIO Y. Neural machine translation by jointly learning to align and translate[J]. arXiv:1409.0473, 2014.
[10] SEE A, LIU P J, MANNING C D. Get to the point: summarization with pointer-generator networks[J]. arXiv:1704. 04368, 2017.
[11] LEWIS M, LIU Y H, GOYAL N, et al. BART: denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension[J]. arXiv:1910. 13461, 2019.
[12] ZHANG J Q, ZHAO Y, SALEH M, et al. PEGASUS: pre-training with extracted gap-sentences for abstractive summarization[J]. arXiv:1912.08777, 2019.
[13] JOSHI A, FIDALGO E, ALEGRE E, et al. DeepSumm: exploiting topic models and sequence to sequence networks for extractive text summarization[J]. Expert Systems with Applications, 2023, 211: 118442.
[14] WANG D Q, LIU P F, ZHENG Y N, et al. Heterogeneous graph neural networks for extractive document summarization[J]. arXiv:2004.12393, 2020.
[15] MAO Q R, ZHU H D, LIU J N, et al. MuchSum: multi-channel graph neural network for extractive summarization[C]//Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval. New York: ACM, 2022: 2617-2622.
[16] ZHONG M, LIU P F, CHEN Y R, et al. Extractive summarization as text matching[J]. arXiv:2004.08795, 2020.
[17] NARAYAN S, COHEN S B, LAPATA M. Ranking sentences for extractive summarization with reinforcement learning[J]. arXiv:1802.08636, 2018.
[18] BIAN J Y, HUANG X D, ZHOU H, et al. GoSum: extractive summarization of long documents by reinforcement learning and graph-organized discourse state[J]. Knowledge and Information Systems, 2024, 66(12): 7557-7580.
[19] CHENG J P, LAPATA M. Neural summarization by extracting sentences and words[J]. arXiv:1603.07252, 2016.
[20] KWON T, LEE S Y. OrderSum: semantic sentence ordering for extractive summarization[J]. arXiv:2502.16180, 2025.
[21] YASUNAGA M, ZHANG R, MEELU K, et al. Graph-based neural multi-document summarization[J]. arXiv:1706.06681, 2017.
[22] JING B Y, YOU Z Y, YANG T, et al. Multiplex graph neural network for extractive text summarization[J]. arXiv:2108. 12870, 2021.
[23] YASUNAGA M, KASAI J, ZHANG R, et al. ScisummNet: a large annotated corpus and content-impact models for scientific paper summarization with citation networks[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2019: 7386-7393.
[24] XU J C, GAN Z, CHENG Y, et al. Discourse-aware neural extractive text summarization[J]. arXiv:1910.14142, 2019.
[25] ONAN A, ALHUMYANI H. Contextual hypergraph networks for enhanced extractive summarization: introducing multi-element contextual hypergraph extractive summarizer (MCHES)[J]. Applied Sciences, 2024, 14(11): 4671.
[26] GONG S, ZHU Z F, QI J T, et al. SeburSum: a novel set-based summary ranking strategy for summary-level extractive summarization[J]. The Journal of Supercomputing, 2023, 79(12): 12949-12977.
[27] BELTAGY I, PETERS M E, COHAN A. Longformer: the long-document transformer[J]. arXiv:2004.05150, 2020.
[28] CHENG X X, SHEN Y L, LU W M. A set prediction network for extractive summarization[C]//Findings of the Association for Computational Linguistics: ACL 2023. Stroudsburg: ACL, 2023: 4766-4777.
[29] GANGUNDI R, SRIDHAR R. RBCA-ETS: enhancing extractive text summarization with contextual embedding and word-level attention[J]. International Journal of Information Technology, 2025, 17(2): 1127-1135.
[30] ZHANG H P, LIU X, ZHANG J W. Extractive summarization via ChatGPT for faithful summary generation[J]. arXiv:2304.04193, 2023.
[31] LIU Y, ITER D, XU Y C, et al. G-Eval: NLG evaluation using GPT-4 with better human alignment[J]. arXiv:2303. 16634, 2023.
[32] CHEN Y C, BANSAL M. Fast abstractive summarization with reinforce-selected sentence rewriting[J]. arXiv:1805. 11080, 2018.
[33] ZHANG T Y, KISHORE V, WU F, et al. BERTScore: evaluating text generation with BERT[J]. arXiv:1904.09675, 2019.
[34] FU J L, NG S K, JIANG Z B, et al. GPTScore: evaluate as you desire[J]. arXiv:2302.04166, 2023.
[35] LIU Y X, LIU P F, RADEV D, et al. BRIO: bringing order to abstractive summarization[J]. arXiv:2203.16804, 2022.
[36] ZHONG Q H, DING L, LIU J H, et al. E2S2: encoding-enhanced sequence-to-sequence pretraining for language understanding and generation[J]. IEEE Transactions on Knowledge and Data Engineering, 2024, 36(12): 8037-8050.
[37] VASWANI A. Attention is all you need[C]//Advances in Neural Information Processing Systems, 2017: 5998-6008.
[38] GU J T, LU Z D, LI H, et al. Incorporating copying mechanism in sequence-to-sequence learning[J]. arXiv:1603. 06393, 2016.
[39] XU S, LI H R, YUAN P, et al. Self-attention guided copy mechanism for abstractive summarization[C]//Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics. Stroudsburg: ACL, 2020: 1355-1362.
[40] RYU S, DO H, KIM Y, et al. Multi-dimensional optimization for text summarization via reinforcement learning[J]. arXiv:2406.00303, 2024.
[41] FU Y, XIONG D Y, DONG Y. Inverse reinforcement learning for text summarization[J]. arXiv:2212.09917, 2022.
[42] VIVEK A, DEVI V S. SumBART—an improved BART model for abstractive text summarization[C]//Proceedings of the International Conference on Neural Information Processing. Singapore: Springer Nature, 2023: 313-323.
[43] COHAN A, DERNONCOURT F, KIM D S, et al. A discourse-aware attention model for abstractive summarization of long documents[J]. arXiv:1804.05685, 2018.
[44] SUN S C, YUAN R F, HE J F, et al. Data selection curriculum for abstractive text summarization[C]//Findings of the Association for Computational Linguistics: EMNLP 2023. Stroudsburg: ACL, 2023: 7990-7995.
[45] KUMAR S, SOLANKI A. An abstractive text summarization technique using transformer model with self-attention mechanism[J]. Neural Computing and Applications, 2023, 35(25): 18603-18622.
[46] CHOWDHURY T, KUMAR S, CHAKRABORTY T. Neural abstractive summarization with structural attention[J]. arXiv:2004.09739, 2020.
[47] ZHAO Y, HUANG S P, ZHOU D S, et al. CNsum: automatic summarization for Chinese news text[C]//Proceedings of the International Conference on Wireless Algorithms, Systems, and Applications. Cham: Springer, 2022: 539-547.
[48] DILAWARI A, KHAN M U G, SALEEM S, et al. Neural attention model for abstractive text summarization using linguistic feature space[J]. IEEE Access, 2023, 11: 23557-23564.
[49] ARGADE D, KHAIRNAR V, VORA D, et al. Multimodal Abstractive Summarization using bidirectional encoder representations from transformers with attention mechanism[J]. Heliyon, 2024, 10(4): e26162.
[50] MINAIDI M N, PAPAIOANNOU C, POTAMIANOS A. Self-attention based generative adversarial networks for unsupervised video summarization[C]//Proceedings of the 2023 31st European Signal Processing Conference. Piscataway: IEEE, 2023: 571-575.
[51] YANG F Y, CUI R Y, YI Z W, et al. Cross-language generative automatic summarization based on attention mechanism[C]//Proceedings of the 17th International Conference on Web Information Systems and Applications. Cham: Springer, 2020: 236-247.
[52] LI Q, WAN W B, ZHAO Y M, et al. Improved BIO-based Chinese automatic abstract-generation model[J]. ACM Transactions on Asian and Low-Resource Language Information Processing, 2024, 23(3): 1-16.
[53] PAULUS R, XIONG C M, SOCHER R. A deep reinforced model for abstractive summarization[J]. arXiv:1705.04304, 2017.
[54] BEKEN FIKRI F, OFLAZER K, YAN?KO?LU B. Abstractive summarization with deep reinforcement learning using semantic similarity rewards[J]. Natural Language Engineering, 2024, 30(3): 554-576.
[55] KENESHLOO Y, RAMAKRISHNAN N, REDDY C K. Deep transfer reinforcement learning for text summarization[C]//Proceedings of the 2019 SIAM International Conference on Data Mining. Philadelphia, PA: Society for Industrial and Applied Mathematics, 2019: 675-683.
[56] DEVLIN J, CHANG M W, LEE K, et al. BERT: pre-training of deep bidirectional transformers for language understanding[J]. arXiv:1810.04805, 2018.
[57] RADFORD A, NARASIMHAN K. Improving language understanding by generative pre-training[EB/OL]. (2018-06?11)[2025?04?16]. https://openai.com/index/language-unsupervised/.
[58] RADFORD A, WU J, CHILD R, et al. Language models are unsupervised multitask learners[J].OpenAI Blog, 2019(8): 9.
[59] RAFFEL C, SHAZEER N, ROBERTS A, et al. Exploring the limits of transfer learning with a unified text-to-text transformer[J]. Journal of Machine Learning Research, 2020, 21(1): 5485-5551.
[60] CHEN T, WANG X W, YUE T W, et al. Enhancing abstractive summarization with extracted knowledge graphs and multi-source transformers[J]. Applied Sciences, 2023, 13(13): 7753.
[61] XU L Y, SU Z L, YU M, et al. Identifying factual inconsistencies in summaries: grounding LLM inference via task taxonomy[C]//Findings of the Association for Computational Linguistics: EMNLP 2024. Stroudsburg: ACL, 2024: 14626-14641.
[62] BALACHANDRAN V, HAJISHIRZI H, COHEN W W, et al. Correcting diverse factual errors in abstractive summarization via post?editing and language model infilling[J]. arXiv:2210.12378, 2022.
[63] REDA A, SALAH N, ADEL J, et al. A hybrid Arabic text summarization approach based on transformers[C]//Proceedings of the 2022 2nd International Mobile, Intelligent, and Ubiquitous Computing Conference. Piscataway: IEEE, 2022: 56-62.
[64] WANG J A, MENG F D, ZHENG D, et al. Towards unifying multi-lingual and cross-lingual summarization[J]. arXiv:2305.09220, 2023.
[65] JIANG P C, XIAO C, WANG Z F, et al. TriSum: learning summarization ability from large language models with structured rationale[J]. arXiv:2403.10351, 2024.
[66] NALLAPATI R, ZHOU B W, DOS SANTOS C N, et al. Abstractive text summarization using sequence-to-sequence RNNs and beyond[J]. arXiv:1602.06023, 2016.
[67] NARAYAN S, COHEN S B, LAPATA M. Don’t give me the details, just the summary! topic-aware convolutional neural networks for extreme summarization[J]. arXiv:1808. 08745, 2018.
[68] HU B T, CHEN Q C, ZHU F Z. LCSTS: a large scale Chinese short text summarization dataset[J]. arXiv:1506.05865, 2015.
[69] 闫晓东, 王羿钦, 黄硕, 等. 藏文文本摘要数据集[J]. 中国科学数据, 2022, 7(2): 43-49.
YAN X D, WANG Y Q, HUANG S, et al. A dataset of Tibetan text summarization[J]. China Scientific Data, 2022, 7(2): 43-49.
[70] 翁彧, 邢天娇, 叶旭明, 等. 中-蒙-藏-维文多文档摘要数据集[J]. 中国科学数据, 2024, 9(4): 92-103.
WENG Y, XING T J, YE X M, et al. A dataset of Chinese-Mongolian-Tibetan-Uyghur multi-document summaries[J]. China Scientific Data, 2024, 9(4): 92-103.
[71] 欧阳新鹏, 闫晓东. 藏汉跨语言摘要数据集TiCLS[J]. 中国科学数据, 2024, 9(4): 75-82.
OUYANG X P, YAN X D. A dataset of Tibetan-Chinese cross-lingual summaries: TiCLS[J]. China Scientific Data, 2024, 9(4): 75-82.
[72] LIN C Y. ROUGE: a package for automatic evaluation of summaries[C]//Proceedings of the Annual Meeting of the Association for Computational Linguistics, 2004: 74-81.
[73] PAPINENI K, ROUKOS S, WARD T, et al. BLEU: a method for automatic evaluation of machine translation[C]//Proceedings of the 40th Annual Meeting on Association for Computational Linguistics. Stroudsburg: ACL, 2001: 311-318.
[74] BANERJEE S, LAVIE A. METEOR: an automatic metric for MT evaluation with improved correlation with human judgments[C]//Proceedings of the ACL Workshop on Intrinsic and Extrinsic Evaluation Measures for Machine Translation and/or Summarization, 2005: 65-72.
[75] SHEN C H, CHENG L Y, NGUYEN X P, et al. Large language models are not yet human-level evaluators for abstractive summarization[J]. arXiv:2305.13091, 2023.
[76] LI L H, ZHANG H P, LI C J, et al. Evaluation on ChatGPT for Chinese language understanding[J]. Data Intelligence, 2023, 5(4): 885-903.
[77] CHEN Y M, CHEN H B, LIU S M, et al. Research on information extraction of LCSTS dataset based on an improved BERTSum-LSTM model[C]//Proceedings of the 2024 2nd International Conference on Mechatronics, IoT and Industrial Informatics. Piscataway: IEEE, 2024: 226-231.
[78] YU T Z, LIU Z H, FUNG P. AdaptSum: towards low-resource domain adaptation for abstractive summarization[J]. arXiv:2103.11332, 2021.
[79] LI Y H, MIAO S Y, HUANG H Y, et al. Word matters: what influences domain adaptation in summarization? [J]. arXiv:2406.14828, 2024.
[80] ?AGAR A, ROBNIK-?IKONJA M. Cross-lingual transfer of abstractive summarizer to less-resource language[J]. Journal of Intelligent Information Systems, 2022, 58(1): 153-173.
[81] ARSLAN M, GHANEM H, MUNAWAR S, et al. A survey on RAG with LLMs[J]. Procedia Computer Science, 2024, 246: 3781-3790.
[82] PU X, GAO M Q, WAN X J. Summarization is (almost) dead[J]. arXiv:2309.09558, 2023.
[83] SUN S C, YUAN R F, CAO Z Q, et al. Prompt chaining or stepwise prompt? refinement in text summarization[C]//Findings of the Association for Computational Linguistics: ACL 2024. Stroudsburg: ACL, 2024: 7551-7558.
[84] LAMAAKAL I, MALEH Y, EL MAKKAOUI K, et al. Tiny language models for automation and control: overview, potential applications, and future research directions[J]. Sensors, 2025, 25(5): 1318.
[85] BADSHAH S, SAJJAD H. Reference-guided verdict: LLMs-as-judges in automatic evaluation of free-form text[J]. arXiv:2408.09235, 2024.
[86] CHEN S Q, GAO S Y, HE J X. Evaluating factual consistency of summaries with large language models[J]. arXiv:2305. 14069, 2023.
[87] PARK G, HWANG S, LEE H. Low-resource cross-lingual summarization through few?shot learning with large language models[J]. arXiv:2406.04630, 2024.
[88] BAI Y, GAO Y, HUANG H Y. Cross?lingual abstractive summarization with limited parallel resources[J]. arXiv:2105.13648, 2021.
[89] LIU Y, LAPATA M. Hierarchical transformers for multi-document summarization[J]. arXiv:1905.13164, 2019.
[90] LI M, QI J Z, LAU J H. Compressed heterogeneous graph for abstractive multi-document summarization[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2023: 13085-13093.