基于平面化句子表示提升关系抽取性能

doi:10.3778/j.issn.1002-8331.2407-0179

摘要/Abstract

摘要： 关系抽取旨在识别句子中命名实体之间的关系。上下文语境决定了命名实体间的真实关系，因此研究者们期望模型能够正确识别上下文语境中反映的实体语义关系。然而，现有的关系抽取模型在很大程度上倾向于对实体名称模式的记忆，即通过学习和匹配特定实体名称组合来推断关系，而非基于上下文信息进行关系判断。这种对上下文信息的忽视可能导致模型在面对复杂语境、多义实体或罕见关系表达时，其预测性能受到限制，无法达到理想状态下的高精度与强泛化。为了解决这一问题，通过将输入文本转化为抽象的语义平面，并直接基于整个语义平面的信息进行关系判断，从而强制模型在预测实体间关系时考虑上下文信息。实验证明，该方法在中文文学文本语料、ACE05英文和SemEval-2010 Task-8三个数据集上的F1分数分别超过之前的相关研究1.9、1.03和0.39个百分点，有效提高了模型对上下文信息的利用程度和对复杂文本中实体关系的理解深度。

关键词: 关系抽取, 语义平面, 实体偏差, 平面化句子表示

Abstract: Relation extraction aims to identify the relationships between named entities within a sentence. The context determines the true relationships between these named entities. Therefore, researchers expect models to correctly recognize the semantic relationships of entities as reflected in the context. However, existing relation extraction models largely tend to memorize patterns of entity names, that is, they infer relationships by learning and matching specific combinations of entity names rather than making relational judgments based on contextual information. This neglect of contextual information may lead to limited predictive performance when models encounter complex contexts, ambiguous entities, or rare relational expressions, failing to achieve high accuracy and strong generalization under ideal conditions. To address this issue, this paper transforms input text into an abstract semantic plane and directly performs relation judgments based on the information from the entire semantic plane, thereby forcing the model to consider contextual information when predicting relationships between entities. Experiments demonstrate that this method surpasses previous related studies by 1.9, 1.03, and 0.39 percentage points in F1 scores on three datasets: Chinese literary text corpus, ACE05 English, and SemEval-2010 Task-8, respectively, effectively enhancing the model‘s utilization of contextual information and its understanding depth of entity relationships in complex texts.

Key words: relation extraction, semantic plane, entity bias, planarized sentence representation

王昊, 陈艳平, 黄瑞章, 秦永彬. 基于平面化句子表示提升关系抽取性能[J]. 计算机工程与应用, 2025, 61(16): 171-178.

WANG Hao, CHEN Yanping, HUANG Ruizhang, QIN Yongbin. Improving Sentence-Level Relation Extraction Through Planarized Sentence Representation[J]. Computer Engineering and Applications, 2025, 61(16): 171-178.

参考文献

[1] ZHONG Z X, CHEN D Q. A frustratingly easy approach for entity and relation extraction[C]//Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. Stroudsburg: ACL, 2021: 50-61.
[2] 李雅杰, 唐国根, 李平. DPMN: 面向重叠关系抽取问题的多任务学习网络[J]. 计算机工程与应用, 2024, 60(20): 160-167.
LI Y J, TANG G G, LI P. DPMN: multi-task learning network for problem of overlapping relation extraction[J]. Computer Engineering and Applications, 2024, 60(20): 160-167.
[3] WANG H, CHEN Y P, YANG W Z, et al. A two dimensional feature engineering method for relation extraction[J]. arXiv:2404.04959, 2024.
[4] REN L, LIU Y B, CAO Y X, et al. CoVariance-based causal debiasing for entity and relation extraction[C]//Findings of the Association for Computational Linguistics: EMNLP 2023. Stroudsburg: ACL, 2023: 2627-2640.
[5] WANG Y W, CHEN M H, ZHOU W X, et al. Should we rely on entity mentions for relation extraction? Debiasing relation extraction with counterfactual analysis[C]//Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. Stroudsburg: ACL, 2022: 3071-3081.
[6] GENG R S, CHEN Y P, HUANG R Z, et al. Planarized sentence representation for nested named entity recognition[J]. Information Processing & Management, 2023, 60(4): 103352.
[7] WANG J B, LUO W Q, YAN X W, et al. A relation extraction model for enhancing subject features and relational attention[C]//Proceedings of the 20th International Conference on Web Information Systems and Applications. Singapore: Springer, 2023: 259-270.
[8] NGUYEN T H, GRISHMAN R. Relation extraction: perspective from convolutional neural networks[C]//Proceedings of the 1st Workshop on Vector Space Modeling for Natural Language Processing. Stroudsburg: ACL, 2015: 39-48.
[9] ZHANG Y H, ZHONG V, CHEN D Q, et al. Position-aware attention and supervised data improve slot filling[C]//Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2017: 35-45.
[10] CHEN Y P, YANG W Z, WANG K, et al. A neuralized feature engineering method for entity relation extraction[J]. Neural Networks, 2021, 141: 249-260.
[11] TIAN Y H, CHEN G M, SONG Y, et al. Dependency-driven relation extraction with attentive graph convolutional networks[C]//Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing. Stroudsburg: ACL, 2021: 4458-4471.
[12] CHEN G M, TIAN Y H, SONG Y, et al. Relation extraction with type-aware map memories of word dependencies[C]// Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021. Stroudsburg: ACL, 2021: 2501-2512.
[13] ZHOU W X, CHEN M H. An improved baseline for sentence-level relation extraction[C]//Proceedings of the 2nd Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the 12th International Joint Conference on Natural Language Processing. Stroudsburg: ACL, 2022: 161-168.
[14] HUANG J Y, LI B Z, XU J S, et al. Unified semantic typing with meaningful label inference[C]//Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. Stroudsburg: ACL, 2022: 2642-2654.
[15] QIN H, TIAN Y H, SONG Y. Relation extraction with word graphs from N-grams[C]//Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2021: 2860-2868.
[16] QIN H, TIAN Y, SONG Y. Enhancing relation extraction via adversarial multi-task learning[C]//Proceedings of the 13th Language Resources and Evaluation Conference, 2022: 6190-6199.
[17] WANG Y W, HOOI B, WANG F, et al. How fragile is relation extraction under entity replacements?[C]//Proceedings of the 27th Conference on Computational Natural Language Learning. Stroudsburg: ACL, 2023: 414-423.
[18] XU J J, WEN J, SUN X, et al. A discourse-level named entity recognition and relation extraction dataset for Chinese literature text[J]. arXiv:1711.07010, 2017.
[19] QIN Y B, YANG W Z, WANG K, et al. Entity relation extraction based on entity indicators[J]. Symmetry, 2021, 13(4): 539.
[20] TIAN Y H, SONG Y, XIA F. Improving relation extraction through syntax-induced pre-training with dependency masking[C]//Findings of the Association for Computational Linguistics: ACL 2022. Stroudsburg: ACL, 2022: 1875-1886.
[21] LI B, YU D Y, YE W, et al. Sequence generation with label augmentation for relation extraction[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2023, 37(11): 13043-13050.
[22] ZHANG D H, LIU Z Y, JIA W Q, et al. Dual attention graph convolutional network for relation extraction[J]. IEEE Transactions on Knowledge and Data Engineering, 2024, 36(2): 530-543.
[23] HENDRICKX I, KIM S N, KOZAREVA Z, et al. SemEval-2010 task 8: multi-way classification of semantic relations between pairs of nominals[J]. arXiv:1911.10422, 2019.
[24] SOCHER R, PENNINGTON J, HUANG E H, et al. Semi-supervised recursive autoencoders for predicting sentiment distributions[C]//Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2011: 151-161.
[25] ZENG D, LIU K, LAI S, et al. Relation classification via convolutional deep neural network[C]//Proceedings of the 25th International Conference on Computational Linguistics: Technical Papers, 2014: 2335-2344.
[26] DOS SANTOS C, XIANG B, ZHOU B W. Classifying relations by ranking with convolutional neural networks[C]//Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing. Stroudsburg: ACL, 2015: 626-634.
[27] XU Y, MOU L L, LI G, et al. Classifying relations via long short term memory networks along shortest dependency paths[C]//Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2015: 1785-1794.
[28] LIU Y, WEI F R, LI S J, et al. A dependency-based neural network for relation classification[C]//Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing. Stroudsburg: ACL, 2015: 285-290.
[29] CAI R, ZHANG X D, WANG H F. Bidirectional recurrent convolutional neural network for relation classification[C]//Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics. Stroudsburg: ACL, 2016: 756-765.
[30] WEN J, SUN X, REN X C, et al. Structure regularized neural network for entity relation classification for Chinese literature text[C]//Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. Stroudsburg: ACL, 2018: 365-370.
[31] CHEN Y P, ZHONG X Y, LIU Y J, et al. A deep penetration network for sentence classification[J]. Information Fusion, 2023, 95: 174-185.
[32] 魏超, 陈艳平, 王凯, 等. 基于掩码提示与门控记忆网络校准的关系抽取方法[J]. 计算机应用, 2024, 44(6): 1713-1719.
WEI C, CHEN Y P, WANG K, et al. Relation extraction method based on mask prompt and gated memory network calibration[J]. Journal of Computer Applications, 2024, 44(6): 1713-1719.
[33] LU Y J, LIU Q, DAI D, et al. Unified structure generation for universal information extraction[J]. arXiv:2203.12277, 2022.
[34] CHEN Y P, ZHENG Q H, CHEN P. A set space model for feature calculus[J]. IEEE Intelligent Systems, 2017, 32(5): 36-42.
[35] XU J L, CHEN Y P, QIN Y B, et al. A feature combination-based graph convolutional neural network model for relation extraction[J]. Symmetry, 2021, 13(8): 1458.
[36] XU J L, CHEN Y P, QIN Y B, et al. A learnable graph convolutional neural network model for relation extraction[C]//Proceedings of the 28th China Conference on Information Retrieval. Cham: Springer, 2022: 90-104.
[37] WU S C, HE Y F. Enriching pre-trained language model with entity information for relation classification[C]//Proceedings of the 28th ACM International Conference on Information and Knowledge Management. New York: ACM, 2019: 2361-2364.
[38] BALDINI SOARES L, FITZGERALD N, LING J, et al. Matching the blanks: distributional similarity for relation learning[C]//Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics. Stroudsburg: ACL, 2019: 2895-2905.
[39] TIAN Y H, SONG Y. Combinatory grammar tells underlying relevance among entities[C]//Findings of the Association for Computational Linguistics: EMNLP 2022. Stroudsburg: ACL, 2022: 5780-5786.