To address the challenges in target-domain knowledge acquisition, such as data sparsity, semantic complexity and rapid dynamic evolution, and the problems of poor domain adaptability and dependence on large-scale annotation in existing knowledge extraction methods, a target domain knowledge extraction method fusing ontology and prompt learning is proposed. Firstly, in order to deal with the professionalism and dynamics of the target domain concept system, a dynamically evolving domain ontology is constructed through a closed-loop iterative mechanism of “expert definition-large language model (LLM) expansion-rule validation”, which provides precise semantic constraints for the knowledge extraction. Secondly, in order to solve the extraction difficulties in the target domain, a hybrid extraction framework based on structured prompts is designed. Through dynamic pattern injection and a parallel Transformer architecture, the efficient joint extraction of entities and relations is achieved. Finally, a curriculum learning strategy is employed to enhance the model’s convergence and generalization capabilities under few-shot conditions. Experimental results demonstrate that the reasoning capabilities of LLM are effectively fused with the precision of domain knowledge by the proposed method, significantly improving the accuracy and robustness of knowledge extraction in the target domain.
LiuY, YaoY S, TonJ F, et al. Trustworthy LLMs: a survey and guideline for evaluating large language models’ alignment[PP/OL]. V2. arXiv (2024-03-21)[2025-10-30].
[4]
ZhaoW X, ZhouK, LiJ Y, et al. A survey of large language models[PP/OL]. V16. arXiv (2025-03-11)[2025-10-30].
TangX Y, HuangY, XiaM, et al. A multi-task BERT-BiLSTM-AM-CRF strategy for Chinese named entity recognition[J]. Neural Processing Letters, 2023,55(2):1209-1229.
[12]
WeiX, CuiX Y, ChengN, et al. ChatIE: zero-shot information extraction via chatting with ChatGPT[PP/OL]. (2023-02-20)[2025-10-30].
[13]
PolakM P, MorganD. Extracting accurate materials data from research papers with conversational language models and prompt engineering[J]. Nature Communications, 2024,15:1569.
[14]
HedderichM A, LangeL, AdelH, et al. A survey on recent approaches for natural language processing in low-resource scenarios[C]//Proceedings of the 2021 conference of the north american chapter of the association for computational linguistics: human language technologies. Stroudsburg, USA: ACL, 2021. DOI: 10.18653/v1/2021.naacl-main.201 .
FritzlerA, LogachevaV, KretovM. Few-shot classification in named entity recognition task[C]//Proceedings of the 34th ACM/SIGAPP symposium on applied computing. New York, USA: ACM, 2019:993-1000.
[17]
HouY T, CheW X, LaiY K, et al. Few-shot slot tagging with collapsed dependency transfer and label-enhanced task-adaptive projection network[C]//Proceedings of the 58th annual meeting of the association for computational linguistics. Stroudsburg, USA: ACL, 2020:1381-1393.
[18]
LiJ, ShangS, ShaoL. MetaNER: named entity recognition with meta-learning[C]//Proceedings of the web conference 2020. New York, USA: ACM, 2020. DOI: 10.1145/3366423.3380127 .
[19]
BapnaA, TürG, Hakkani-TürD, et al. Towards zero-shot frame semantic parsing for domain scaling[C]//Proceedings of the interspeech 2017. Baixas, France: ISCA, 2017:2476-2480.
[20]
LiuZ H, WinataG I, FungP. Zero-resource cross-domain named entity recognition[C]//Proceedings of the 5th workshop on representation learning for NLP. Stroudsburg, USA: ACL, 2020. DOI: 10.18653/v1/2020.repl4nlp-1.1 .
[21]
CuiL Y, WuY, LiuJ, et al. Template-based named entity recognition using BART[C]//Findings of the association for computational linguistics: ACL-IJCNLP 2021. Stroudsburg, USA: ACL, 2021. DOI: 10.18653/v1/2021.findings-acl.161 .
[22]
MaR T, ZhouX, GuiT, et al. Template-free prompt tuning for few-shot NER[PP/OL]. V3. arXiv (2022-11-23)[2025-10-30].
[23]
LiD F, HuB T, ChenQ C. Prompt-based text entailment for low-resource named entity recognition[C]//Proceedings of the 29th international conference on computational linguistics. Stroudsburg, USA: ACL, 2022:1896-1903.
[24]
DevlinJ, ChangM W, LeeK, 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. Stroudsburg, USA: ACL, 2019:4171-4186.
[25]
RobertsA, RaffelC, ShazeerN. How much knowledge can you pack into the parameters of a language model?[C]//Proceedings of the 2020 conference on empirical methods in natural language processing. Stroudsburg, USA: ACL, 2020. DOI: 10.18653/v1/2020.emnlp-main.437 .