基于大语言模型的公安专业小样本知识抽取方法研究

doi:10.3778/j.issn.1673-9418.2403039

摘要/Abstract

摘要： 当前公安业务工作信息化、数字化飞速发展，在公安工作中产生了大量执法办案数据，但是其文本种类较多、信息量较大，导致一线民警在阅卷工作中常面临阅读效率低、信息难以聚合等问题。为更进一步利用执法办案文本，需要对其进行智能分析、知识抽取，但受限于公安专业执法办案文本的专业性、数据敏感性、保密性，以及公安数据出网要求等，仅能获取到少量学习训练样本，使用传统的深度学习模型抽取效果不尽如人意。因此提出使用较少资源和数据构建垂直领域大语言模型，实现模型对公安专业适配的方法，利用知识编辑技术MEMIT、低资源微调技术LoRA、提示模板，提高模型对警务术语、警务常识等公安知识的理解能力。为进一步提高模型的知识抽取效果，设计小样本执法办案文本数据抽取流程，以更好结合模型中的相关案别专业知识。实验结果表明，融合抽取流程的公安专业垂直领域大语言模型在各类知识抽取任务中准确率较之传统方法显著提高，有助于帮助一线民警快速、客观、准确分析执法办案文本，挖掘案件潜在信息，支撑公安工作智能化发展。

关键词: 大语言模型, 知识抽取, 小样本数据, 公安执法办案

Abstract: The rapid development of informatization and digitalization in public security business has generated a large amount of law enforcement case data in public security work. However, due to various types of text and large amount of information, front-line police officers often face problems such as low reading efficiency and difficulty in aggregating information in the process of reading case files. In order to further utilize the law enforcement case text, it is necessary to conduct intelligent analysis and knowledge extraction. However, due to the professionalism, data sensitivity, confidentiality of public security professional law enforcement case text, as well as the requirements of public security data going out of the network, only a small number of learning training samples can be obtained, and the traditional deep learning model has unsatisfactory extraction effect. Therefore, this paper proposes to build a large language model in vertical fields with fewer resources and data, and realize the adaptation of the model to the public security profession. The model uses knowledge editing technology MEMIT (mess-editing memory in a transformer), low-resource fine-tuning technology LoRA (low-rank adaptation), and prompt templates to improve the model??s understanding of public security knowledge such as police terminology and common sense. Moreover, in order to further improve the knowledge extraction effect of the model, a small sample law enforcement case text data extraction process is designed to better integrate the professional knowledge related to the case in the model. Experimental results show that the accuracy of the public security professional vertical field large language model integrated with the extraction process in various knowledge extraction tasks is significantly improved compared with the traditional methods, which helps front-line police officers quickly, objectively and accurately analyze law enforcement case text, dig out potential case information, and support the intelligent development of public security work.

Key words: large language model, knowledge extraction, small sample data, public security law enforcement

裴炳森, 李欣, 蒋章涛, 刘明帅. 基于大语言模型的公安专业小样本知识抽取方法研究[J]. 计算机科学与探索, 2024, 18(10): 2630-2642.

PEI Bingsen, LI Xin, JIANG Zhangtao, LIU Mingshuai. Research on Public Security Professional Small Sample Knowledge Extraction Method Based on Large Language Model[J]. Journal of Frontiers of Computer Science and Technology, 2024, 18(10): 2630-2642.

参考文献

[1] 卢经纬, 郭超, 戴星原, 等. 问答ChatGPT之后:超大预训练模型的机遇和挑战[J]. 自动化学报, 2023, 49(4): 705-717.
LU J W, GUO C, DAI X Y, et al. The ChatGPT after: opportunities and challenges of very large scale pre-trained models[J]. Acta Automatica Sinica, 2023, 49(4): 705-717.
[2] 桑基韬, 于剑. 从ChatGPT看AI未来趋势和挑战[J]. 计算机研究与发展, 2023, 60(6): 1191-1201.
SANG J T, YU J. ChatGPT: a glimpse into AI??s future[J]. Journal of Computer Research and Development, 2023, 60(6): 1191-1201.
[3] 陈舒梦. 大语言模型在外语教学中的应用研究[J]. 长春师范大学学报, 2023, 42(11): 170-173.
CHEN S M. Research on the application of large models in foreign language teaching[J]. Journal of Changchun Normal University, 2023, 42(11): 170-173.
[4] 杨涛, 王欣宇, 朱垚, 等. 大语言模型驱动的中医智能诊疗研究思路与方法[J]. 南京中医药大学学报, 2023, 39(10): 967-971.
YANG T, WANG X Y, ZHU Y, et al. Research ideas and methods of intelligent diagnosis and treatment of traditional Chinese medicine driven by large language model[J]. Journal of Nanjing University of Traditional Chinese Medicine, 2023, 39(10): 967-971.
[5] 杨波, 孙晓虎, 党佳怡, 等. 面向医疗问答系统的大语言模型命名实体识别方法[J]. 计算机科学与探索, 2023, 17(10): 2389-2402.
YANG B, SUN X H, DANG J Y, et al. Named entity recognition method of large language model for medical question answering system[J]. Journal of Frontiers of Computer Science and Technology, 2023, 17(10): 2389-2402.
[6] 王昀, 胡珉, 塔娜, 等. 大语言模型及其在政务领域的应用 [J]. 清华大学学报(自然科学版), 2024, 64(4): 649-658.
WANG Y, HU M, TA N, et al. Large language models and their application in government affairs[J]. Journal of Tsinghua University(Science and Technology), 2024, 64(4): 649-658.
[7] 徐月梅, 胡玲, 赵佳艺, 等. 大语言模型的技术应用前景与风险挑战[J]. 计算机应用, 2024, 44(6): 1655-1662.
XU Y M, HU L, ZHAO J Y, et al. Technology application prospects and risk challenges of large language model[J]. Journal of Computer Applications, 2024, 44(6): 1655-1662.
[8] 祁鹏年, 廖雨伦, 覃飙. 基于深度学习的中文命名实体识别研究综述[J]. 小型微型计算机系统, 2023, 44(9): 1857-1868.
QI P N, LIAO Y L, QIN B. Survey on deep learning for Chinese named entity recognition[J]. Journal of Chinese Computer Systems, 2023, 44(9): 1857-1868.
[9] XUE M G, YU B, LIU T, et al. Porous lattice transformer encoder for Chinese NER[C]//Proceedings of the 28th International Conference on Computational Linguistics, Barcelona, Dec 8-13, 2020: 3831-3841.
[10] CAO P, CHEN Y, LIU K, et al. Adversarial transfer learning for Chinese named entity recognition with self-attention mechanism[C]//Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2018: 182-192.
[11] LI J, FEI H, LIU J, et al. Unified named entity recognition as word-word relation classification[C]//Proceedings of the 2022 AAAI Conference on Artificial Intelligence. Menlo Park: AAAI, 2022: 10965-10973.
[12] QI P, QIN B. SSMI: semantic similarity and mutual information maximization based enhancement for Chinese NER[C]//Proceedings of the 2023 AAAI Conference on Artificial Intelligence. Menlo Park: AAAI, 2023: 13474-13482.
[13] FRITZLER A, LOGACHEVA V, KRETOV M. Few-shot classification in named entity recognition task[C]//Proceedings of the 34th ACM/SIGAPP Symposium on Applied Computing. New York: ACM, 2019: 993-1000.
[14] WANG P, XU R, LIU T, et al. An enhanced span-based decomposition method for few-shot sequence labeling[EB/OL]. [2024-01-14]. https://arxiv.org/abs/2109.13023.
[15] MA R, ZHOU X, GUI T, et al. Template-free prompt tuning for few-shot NER[EB/OL]. [2024-01-14]. https://arxiv.org/abs/2109.13532.
[16] 刘蓓, 许卓明, 陶皖, 等. 少样本关系抽取研究综述[J]. 计算机工程与应用, 2023, 59(15): 27-37.
LIU B, XU Z M, TAO W, et al. Survey on few-shot relation extraction[J]. Computer Engineering and Applications, 2023, 59(15): 27-37.
[17] NAYAK T, NG H T. Effective modeling of encoder-decoder architecture for joint entity and relation extraction[C]//Proceedings of the 2020 AAAI Conference on Artificial Intelligence. Menlo Park: AAAI, 2020: 8528-8535.
[18] XUE F, SUN A, ZHANG H, et al. GDPNet: refining latent multi-view graph for relation extraction[C]//Proceedings of the 2021 AAAI Conference on Artificial Intelligence. Menlo Park: AAAI, 2021: 14194-14202.
[19] YANG S, ZHANG Y, NIU G, et al. Entity concept-enhanced few-shot relation extraction[EB/OL]. [2024-01-14]. https://arxiv.org/abs/2106.02401.
[20] XIE Y, XU H, LI J, et al. Heterogeneous graph neural networks for noisy few-shot relation classification[J]. Knowledge-Based Systems, 2020, 194: 105548.
[21] ZHANG P, LU W. Better few-shot relation extraction with label prompt dropout[EB/OL]. [2024-01-14]. https://arxiv.org/abs/2210.13733.
[22] HE K, HUANG Y, MAO R, et al. Virtual prompt pre-training for prototype-based few-shot relation extraction[J]. Expert Systems with Applications, 2023, 213: 118927.
[23] 刘涛, 蒋国权, 刘姗姗, 等. 低资源场景事件抽取研究综述 [J]. 计算机科学, 2024, 51(2): 217-237.
LIU T, JIANG G Q, LIU S S, et al. Survey of event extraction in low-resource scenarios[J]. Computer Science, 2024, 51(2): 217-237.
[24] 李培峰, 周国栋, 朱巧明. 基于语义的中文事件触发词抽取联合模型[J]. 软件学报, 2016, 27(2): 280-294.
LI P F, ZHOU G D, ZHU Q M. Semantics-based joint model of Chinese event trigger extraction[J]. Journal of Software, 2016, 27(2): 280-294.
[25] 仲伟峰, 杨航, 陈玉博, 等. 基于联合标注和全局推理的篇章级事件抽取[J]. 中文信息学报, 2019, 33(9): 88-95.
ZHONG W F, YANG H, CHEN Y B, et al. Document-level event extraction based on joint labeling and global reasoning[J]. Journal of Chinese Information Processing, 2019, 33(9): 88-95.
[26] 朱培培, 王中卿, 李寿山, 等. 基于篇章信息和Bi-GRU的中文事件检测[J]. 计算机科学, 2020, 47(12): 233-238.
ZHU P P, WANG Z Q, LI S S, et al. Chinese event detection based on document information and Bi-GRU[J]. Computer Science, 2020, 47(12): 233-238.
[27] LIU X, HUANG H, SHI G, et al. Dynamic prefix-tuning for generative template-based event extraction[EB/OL]. [2024-01-14]. https://arxiv.org/abs/2205.06166.
[28] SNELL J, SWERSKY K, ZEMEL R. Prototypical networks for few-shot learning[C]//Advances in Neural Information Processing Systems 30, Long Beach, Dec 4-9, 2017: 4077-4087.