大语言模型驱动下基于零样本语境联想的意图分类

陶汉卿; 程玉虎; 王雪松; 王军

doi:10.12263/DZXB.20250863

您当前的位置：

首页 >

文章列表页 >

大语言模型驱动下基于零样本语境联想的意图分类

学术论文 | 更新时间：2026-06-04

- 大语言模型驱动下基于零样本语境联想的意图分类
- An Intent Classification Method Based on Zero-Shot Context Association Driven by Large Language Models
- 电子学报 2026年54卷第1期页码：219-233
- 作者机构：
  
  中国矿业大学信息与控制工程学院，江苏徐州 221116
- 作者简介：
  
  [ "陶汉卿男，1995年3月出生于安徽省宿州市。现为中国矿业大学信息与控制工程学院助理研究员。主要研究方向为人工智能、数据挖掘和自然语言处理。E-mail: hqtao@cumt.edu.cn" ]
  [ "程玉虎男，1973年8月出生于安徽省淮南市。现为中国矿业大学信息与控制工程学院教授、博士生导师。主要研究方向为强化学习、具身智能。E-mail: chengyuhu@163.com" ]
  [ "王雪松女，1974年12月出生于安徽省泗县。现为中国矿业大学信息与控制工程学院教授、博士生导师。主要研究方向为机器学习、人工智能。中国电子学会会员编号：E190006839S。E-mail: wangxuesongcumt@163.com" ]
  [ "王军男，1981年1月出生于江苏省徐州市。现为中国矿业大学信息与控制工程学院教授、博士生导师。主要研究方向为智能机器人与无人系统、生物特征识别、机器视觉。中国电子学会会员编号：E190089908M。E-mail: jrobot@126.com" ]
- 基金信息：
  
  中国博士后科学基金(2024M753519;GZC20241922);江苏省卓越博士后资助项目(2024ZB721)
- DOI：10.12263/DZXB.20250863
  中图分类号： TP181;TP183
- 收稿：2025-09-28，
  
  录用：2026-01-06，
  
  纸质出版：2026-01-25
- 稿件说明：
移动端阅览
陶汉卿, 程玉虎, 王雪松, 等. 大语言模型驱动下基于零样本语境联想的意图分类[J]. 电子学报, 2026, 54(01): 219-233.

TAO Hanqing, CHENG Yuhu, WANG Xuesong, et al. An Intent Classification Method Based on Zero-Shot Context Association Driven by Large Language Models[J]. Acta Electronica Sinica, 2026, 54(01): 219-233.
陶汉卿, 程玉虎, 王雪松, 等. 大语言模型驱动下基于零样本语境联想的意图分类[J]. 电子学报, 2026, 54(01): 219-233. DOI：10.12263/DZXB.20250863

TAO Hanqing, CHENG Yuhu, WANG Xuesong, et al. An Intent Classification Method Based on Zero-Shot Context Association Driven by Large Language Models[J]. Acta Electronica Sinica, 2026, 54(01): 219-233. DOI：10.12263/DZXB.20250863

摘要

意图分类是自然语言处理领域中的一项基础而关键的任务，其目标在于准确识别用户输入语句所表达的潜在意图，是对话系统、智能客服与人机交互等应用的重要技术支撑。近年来，基于深度学习的意图分类方法取得了显著进展，但其性能高度依赖大规模标注语料与稳定的领域分布，在实际应用中仍面临诸多挑战。尤其在短文本信息稀疏、标签语义抽象以及领域先验不足等低资源情境下，用户表达往往具有信息密度低、语义依赖隐含、表述方式多样等特点；同时，意图标签本身通常具有高度抽象性，不同标签之间语义边界模糊，现有模型难以仅凭文本内部的字面特征充分刻画深层语义与语境关联，进而制约了意图分类模型在低资源与跨场景条件下的泛化能力与鲁棒性。针对上述问题，本文从语义扩展与语境建模的角度出发，尝试突破传统监督学习对显式标注样本与表层字面特征的依赖。不同于将任务直接设定为零样本意图分类，本文在有监督学习框架下引入大语言模型的零样本语境联想能力，利用其蕴含的丰富世界知识与语义推理能力，扩展可学习的语义空间，从而弥补文本信息稀疏与标签语义不足所带来的建模缺陷。基于这一思路，本文提出一种基于大语言模型的零样本语境联想模型（LLM-based Zero-shot Context Association Model，L-ZCAM）。该模型通过构造结构化提示词，引导大语言模型从联想意图与标签定义两个互补视角生成与输入语句相关的补充性语境语义信息，实现文本内部特征与文本外部知识的联合挖掘，并对意图标签的语义内涵进行显式增强。在模型结构设计上，L-ZCAM采用多路特征编码与交叉注意力机制，对原始文本特征、联想语义特征及标签语义特征进行深度交互建模；同时，引入约束引导的联合损失函数，对联想语义与标签语义之间的一致性进行约束，以缓解语义噪声带来的干扰，实现文本内外信息的有效对齐。通过上述设计，L-ZCAM能够更好地感知多义模糊、标签抽象以及表达多样等复杂语境下的语义关联关系，从而提升意图判别的准确性与稳定性。实验结果表明，在CLINC150、Banking77和HWU64三个公开数据集上，L-ZCAM的宏平均F1分数分别较当前最新方法提升2.25%、1.28%和1.29%，在不同任务场景下具有更强的泛化能力与鲁棒性。

Abstract

Intent classification is a fundamental and critical task in natural language processing

aiming to accurately identify the underlying intentions expressed in user utterances. It serves as an essential technical foundation for dialogue systems

intelligent customer service

and human-computer interaction. In recent years

deep learning-based approaches have achieved remarkable progress in intent classification; however

their performance heavily relies on large-scale annotated corpora and stable domain distributions

which poses significant challenges in real-world applications. In low-resource scenarios characterized by sparse short-text information

abstract label semantics

and insufficient domain prior knowledge

user expressions often exhibit low information density

implicit semantic dependencies

and diverse surface forms. Meanwhile

intent labels are typically highly abstract with blurred semantic boundaries

making it difficult for existing models to capture deep semantic representations and contextual associations solely from literal textual features. These issues severely limit the generalization ability and robustness of intent classification models under low-resource and cross-domain settings. To address these challenges

this paper explores intent classification from the perspective of semantic expansion and contextual modeling

aiming to reduce the reliance of traditional supervised learning methods on explicit annotations and shallow lexical features. Unlike approaches that directly formulate the task as zero-shot intent classification

we introduce the zero-shot contextual association capability of large language models into a supervised learning framework. By leveraging the rich world knowledge and semantic reasoning ability encoded in LLMs

the proposed approach expands the learnable semantic space

thereby alleviating the modeling limitations caused by sparse textual information and insufficient label semantics. Based on this idea

we propose an LLM-based zero-shot context association model (L-ZCAM). The model constructs structured prompts to guide LLMs to generate complementary contextual semantic information related to the input utterance from two complementary perspectives: associative intents and label definitions. This design enables joint mining of in-text features and out-of-text knowledge while explicitly enhancing label semantics. From a structural perspective

L-ZCAM adopts multi-branch feature encoders and a cross-attention mechanism to deeply model the interactions among original textual features

associative semantic features

and label semantic features. In addition

a constraint-guided joint loss function is introduced to enforce semantic consistency between associative semantics and label semantics

mitigating the impact of semantic noise and achieving effective alignment between internal and external information. Through these designs

L-ZCAM is able to better capture semantic associations under complex contexts involving polysemy

abstract labels

and diverse expressions

thereby improving the accuracy and stability of intent prediction. Experimental results on three public datasets

i.e.

CLINC150

Banking77

and HWU64

demonstrate that L-ZCAM outperforms state-of-the-art methods by 2.25%

1.28%

and 1.29% in terms of macro-averaged F1 score

respectively

exhibiting stronger generalization ability and robustness across different task scenarios.

关键词

Keywords

references

Weld H , Huang Xiaoqi , Long Siqu , et al . A survey of joint intent detection and slot filling models in natural language understanding [J ] . ACM Computing Surveys , 2023 , 55 ( 8 ): 1 - 38 . DOI: 10.1145/3547138 http://dx.doi.org/10.1145/3547138

Hrycyk L , Zarcone A , Hahn L . Not so fast, classifier-accuracy and entropy reduction in incremental intent classification [C ] // Proceedings of the 3rd Workshop on Natural Language Processing for Conversational AI . Stroudsburg : ACL , 2021 : 52 - 67 . DOI: 10.18653/v1/2021.nlp4convai-1.6 http://dx.doi.org/10.18653/v1/2021.nlp4convai-1.6

Wang Haifeng , Li Jiwei , Wu Hua , et al . Pre-trained language models and their applications [J ] . Engineering , 2023 , 25 : 51 - 65 . DOI: 10.1016/j.eng.2022.04.024 http://dx.doi.org/10.1016/j.eng.2022.04.024

Ferré P , Fraga I , Hinojosa J A . The interplay between language and emotion: A narrative review [J ] . Cognition and Emotion , 2025 , 39 ( 7 ): 1418 - 1445 . DOI: 10.1080/02699931.2025.2549965 http://dx.doi.org/10.1080/02699931.2025.2549965

Wang Jindong , Lan Cuiling , Liu Chang , et al . Generalizing to unseen domains: A survey on domain generalization [J ] . IEEE Transactions on Knowledge and Data Engineering , 2023 , 35 ( 8 ): 8052 - 8072 . DOI: 10.1109/TKDE.2022.3178128 http://dx.doi.org/10.1109/TKDE.2022.3178128

Gardazi N M , Daud A , Malik M K , et al . BERT applications in natural language processing: A review [J ] . Artificial Intelligence Review , 2025 , 58 ( 6 ): 166 . DOI: 10.1007/s10462-025-11162-5 http://dx.doi.org/10.1007/s10462-025-11162-5

Tompkins V , Montgomery D E , Dore R A , et al . Theory of mind and text comprehension across the lifespan: A meta-analysis [J ] . Developmental Psychology , 2025 , 61 ( 6 ): 1112 - 1125 . DOI: 10.1037/dev0001869 http://dx.doi.org/10.1037/dev0001869

Bzdok D , Thieme A , Levkovskyy O , et al . Data science opportunities of large language models for neuroscience and biomedicine [J ] . Neuron , 2024 , 112 ( 5 ): 698 - 717 . DOI: 10.1016/j.neuron.2024.01.016 http://dx.doi.org/10.1016/j.neuron.2024.01.016

OpenAI , Achiam J , Adler S , et al . GPT-4 technical report [PP/OL ] . V6.arXiv ( 2024-03-04 )[ 2026-01-04 ] . https://doi.org/10.48550/arXiv.2303.08774 https://doi.org/10.48550/arXiv.2303.08774 .

Anil R , Dai A M , Firat O , et al . PaLM 2 technical report [PP/OL ] . V3.arXiv ( 2023-09-13 )[ 2026-01-04 ] . https://doi.org/10.48550/arXiv.2305.10403 https://doi.org/10.48550/arXiv.2305.10403 .

Touvron H , Lavril T , Izacard G , et al . LLaMA: Open and efficient foundation language models [PP/OL ] . V1.arXiv ( 2023-02-27 )[ 2026-01-04 ] . https://doi.org/10.48550/arXiv.2302.13971 https://doi.org/10.48550/arXiv.2302.13971 .

Chang Yupeng , Wang Xu , Wang Jindong , et al . A survey on evaluation of large language models [J ] . ACM Transactions on Intelligent Systems and Technology , 2024 , 15 ( 3 ): 1 - 45 . DOI: 10.1145/3641289 http://dx.doi.org/10.1145/3641289

赵健程 , 冯良骏 , 岳嘉祺 , 等 . 从零样本学习理论模型到工业应用: 动机、演变与挑战 [J ] . 控制与决策 , 2024 , 39 ( 9 ): 2833 - 2857 .

Zhao Jiancheng , Feng Liangjun , Yue Jiaqi , et al . From zero-shot learning theoretical model to its industrial application: Motivation, evolution and challenges [J ] . Control and Decision , 2024 , 39 ( 9 ): 2833 - 2857 . (in Chinese)

Shah C , White R , Andersen R , et al . Using large language models to generate, validate, and apply user intent taxonomies [J ] . ACM Transactions on the Web , 2025 , 19 ( 3 ): 1 - 29 . DOI: 10.1145/3732294 http://dx.doi.org/10.1145/3732294

Zhang Jiong , Chang Weicheng , Yu H F , et al . Fast multi-resolution transformer fine-tuning for extreme multi-label text classification [J ] . Advances in Neural Information Processing Systems , 2021 , 34 : 7267 - 7280 .

Saha Roy R , Katare R , Ganguly N , et al . Discovering and understanding word level user intent in Web search queries [J ] . Journal of Web Semantics , 2015 , 30 : 22 - 38 . DOI: 10.1016/j.websem.2014.07.010 http://dx.doi.org/10.1016/j.websem.2014.07.010

Devlin J , Chang M W , Lee K , et al . BERT: Pre-training of deep bidirectional transformers for language understanding [C ] // Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) . Kerrville : Association for Computational Linguistics 2019 : 4171 - 4186 . DOI: 10.18653/v1/n19-1423 http://dx.doi.org/10.18653/v1/n19-1423

许婷 , 肖桐 , 张圣林 , 等 . 基于LLM的日志故障诊断 [J ] . 电子学报 , 2025 , 53 ( 4 ): 1123 - 1141 .

Xu Ting , Xiao Tong , Zhang Shenglin , et al . Log fault diagnosis based on large language models [J ] . Acta Electronica Sinica , 2025 , 53 ( 4 ): 1123 - 1141 . (in Chinese)

Wang Xiang , Huang Tinglin , Wang Dingxian , et al . Learning intents behind interactions with knowledge graph for recommendation [C ] // Proceedings of the Web Conference 2021 . New York : ACM , 2021 : 878 - 887 . DOI: 10.1145/3442381.3450133 http://dx.doi.org/10.1145/3442381.3450133

吴天舒 , 尹宏鹏 , 赵丹丹 , 等 . 基于迁移学习的零样本故障诊断 [J ] . 电子学报 , 2023 , 51 ( 9 ): 2572 - 2577 .

Wu Tianshu , Yin Hongpeng , Zhao Dandan , et al . Zero sample fault diagnosis based on transfer learning [J ] . Acta Electronica Sinica , 2023 , 51 ( 9 ): 2572 - 2577 . (in Chinese)

Reimers N , Gurevych I . Sentence-BERT: Sentence embeddings using Siamese BERT-networks [C ] // Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing . Stroudsburg : ACL , 2019 : 3980 - 3990 . DOI: 10.18653/v1/d19-1410 http://dx.doi.org/10.18653/v1/d19-1410

Mohsenimofidi S , Prasad A S R , Zahid A , et al . Classifying user intent for effective prompt engineering: A case of a chatbot for startup teams [M ] //Nguyen-Duc A, Abrahamsson P, Khomh F. Generative AI for Effective Software Development . Cham : Springer Nature Switzerland , 2024 : 317 - 329 . DOI: 10.1007/978-3-031-55642-5_15 http://dx.doi.org/10.1007/978-3-031-55642-5_15

Hu E J , Shen Y , Wallis P , et al . Lora: Low-rank adaptation of large language models [C/OL ] // Proceedings of the International Conference on Learning Representations , 2022 : 1 - 13 . https://openreview.net/forum?id=nZeVKeeFYf9 https://openreview.net/forum?id=nZeVKeeFYf9 .

Tian Y J , Han Y K , Chen X S , et al . Beyond answers: Transferring reasoning capabilities to smaller LLMs using multi-teacher knowledge distillation [C ] // Proceedings of the Eighteenth ACM International Conference on Web Search and Data Mining . New York : ACM , 2025 : 251 - 260 . DOI: 10.1145/3701551.3703577 http://dx.doi.org/10.1145/3701551.3703577

Rodriguez J A , Botzer N , Vazquez D , et al . IntentGPT: Few-shot intent discovery with large language models [C ] // Proceedings of the ICLR 2024 Workshop on Large Language Model Agents , 2024 . https://openreview.net/forum?id=IDuQtpSgGp https://openreview.net/forum?id=IDuQtpSgGp .

Dong Kaifang , Jiang Baoxing , Li Hongye , et al . Meta-learning triplet contrast network for few-shot text classification [J ] . Knowledge-Based Systems , 2024 , 303 : 112440 . DOI: 10.1016/j.knosys.2024.112440 http://dx.doi.org/10.1016/j.knosys.2024.112440

Park G , Baek I , Kim B , et al . Dynamic label name refinement for few-shot dialogue intent classification [C ] // Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics . Stroudsburg : ACL , 2025 : 41 - 52 . DOI: 10.18653/v1/2025.acl-short.3 http://dx.doi.org/10.18653/v1/2025.acl-short.3

Wu Mingmin , Hu Yuxue , Zhang Yongcheng , et al . Mitigating idiom inconsistency: A multi-semantic contrastive learning method for Chinese idiom reading comprehension [J ] . Proceedings of the AAAI Conference on Artificial Intelligence , 2024 , 38 ( 17 ): 19243 - 19251 . DOI: 10.1609/aaai.v38i17.29893 http://dx.doi.org/10.1609/aaai.v38i17.29893

Wong B I , Lecompte M , Yang L X . Associative memory with value-directed learning in younger and older adults [J ] . Aging, Neuropsychology, and Cognition , 2025 , 32 ( 6 ): 891 - 906 . DOI: 10.1080/13825585.2025.2539516 http://dx.doi.org/10.1080/13825585.2025.2539516

Connell L , Lynott D . What can language models tell us about human cognition? [J ] . Current Directions in Psychological Science , 2024 , 33 ( 3 ): 181 - 189 . DOI: 10.1177/09637214241242746 http://dx.doi.org/10.1177/09637214241242746

Xu Miao , Li Yufeng , Zhou Zhihua . Robust multi-label learning with PRO loss [J ] . IEEE Transactions on Knowledge and Data Engineering , 2020 , 32 ( 8 ): 1610 - 1624 . DOI: 10.1109/tkde.2019.2908898 http://dx.doi.org/10.1109/tkde.2019.2908898

王进 , 刘彬 , 孙开伟 , 等 . 基于标签关联的多标签演化超网络 [J ] . 电子学报 , 2018 , 46 ( 4 ): 1012 - 1018 .

Wang Jin , Liu Bin , Sun Kaiwei , et al . Multi-label evolutionary hypernetwork based on label correlations [J ] . Acta Electronica Sinica , 2018 , 46 ( 4 ): 1012 - 1018 . (in Chinese)

Larson S , Mahendran A , Peper J J , et al . An evaluation dataset for intent classification and out-of-scope prediction [C ] // Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing . Stroudsburg : ACL , 2019 : 1311 - 1316 . DOI: 10.18653/v1/d19-1131 http://dx.doi.org/10.18653/v1/d19-1131

Coope S , Farghly T , Gerz D , et al . Span-ConveRT: Few-shot span extraction for dialog with pretrained conversational representations [C ] // Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics . Stroudsburg : ACL , 2020 : 107 - 121 . DOI: 10.18653/v1/2020.acl-main.11 http://dx.doi.org/10.18653/v1/2020.acl-main.11

Liu Xingkun , Eshghi A , Swietojanski P , et al . Benchmarking natural language understanding services for building conversational agents [M ] //Marchi E, Siniscalchi S M, Cumani S, et al. Increasing Naturalness and Flexibility in Spoken Dialogue Interaction: 10th International Workshop on Spoken Dialogue Systems . Singapore : Springer Singapore , 2021 : 165 - 183 . DOI: 10.1007/978-981-15-9323-9_15 http://dx.doi.org/10.1007/978-981-15-9323-9_15

Zhao Chao , Vijjini A , Chaturvedi S . PARROT: Zero-shot narrative reading comprehension via parallel reading [C ] // Findings of the Association for Computational Linguistics: EMNLP 2023 . Stroudsburg : ACL , 2023 : 13413 - 13424 . DOI: 10.18653/v1/2023.findings-emnlp.895 http://dx.doi.org/10.18653/v1/2023.findings-emnlp.895

Wang Tianming , Wan Xiaojun . T-CVAE: Transformer-based conditioned variational autoencoder for story completion [C ] // Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence . International Joint Conferences on Artificial Intelligence Organization , 2019 : 5233 - 5239 . DOI: 10.24963/ijcai.2019/727 http://dx.doi.org/10.24963/ijcai.2019/727

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

基于信息融合的区块链系统隐匿安全补丁识别及迁移技术

基于大语言模型语义增强的多模态智能合约漏洞检测方法研究

基于大语言模型的Web文本输入组件测试方法

面向户外多声源增强的鲁棒节点特定分布式广义旁瓣对消

基于超表面的低副瓣高口径效率反射阵列天线