实体关系抽取是自然语言处理的关键任务之一,在藏医药领域的应用对于构建藏医药知识图谱、智能辅助诊断和药物研发具有重要意义。针对藏文医疗文本实体关系抽取任务,提出一种基于预训练模型TiBERT加多头注意力的联合抽取方法。该方法通过TiBERT模型对藏文医疗文本进行编码处理,生成包含上下文信息的特征向量,再利用多头注意力机制增强特征表示能力,捕捉不同实体之间的关联信息。实验结果表明,该模型在藏医文本数据集上的F1值达到81.81%,显著优于其他对比模型,证明了其有效性。
Abstract
Entity relation extraction is one of the key tasks of natural language processing, and its application in the field of Tibetan medicine is of great significance for the construction of a Tibetan medicine knowledge map, intelligent assisted diagnosis and drug research and development. Aiming at the entity relation extraction task of Tibetan medical texts, this paper proposes a joint extraction method based on pre-trained model TiBERT plus multi-head attention. The TiBERT model is used to encode Tibetan medical texts and generate feature vectors containing contextual information. Multi-head attention mechanism is used to enhance feature representation and capture correlation information between different entities. The experimental results show that the F1 value of the model on the Tibetan medicine text dataset reaches 81.81%, which is significantly better than other comparison models, proving its effectiveness.
关键词
TiBERT模型 /
Multi-Head Attention /
实体关系抽取 /
自然语言处理
Key words
TiBERT model /
Multi-Head Attention /
entity relation extraction /
natural language processing
中图分类号：
TP183
TP391.1
R318
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] Humphreys R G,Azzam S,Huyck C,et al.Description of the LaSIE-II System as Used for MUC7[J].1998.
[2] Zeng D,Liu K,Lai S,et al.Relation classification via convolutional deep neural network[C]//International Conference on Computational Linguistics.the 25th International Conference on Computational Linguistics,2014.
[3] Miwa M,Bansal M.End-to-End Relation Extraction using LSTMs on Sequences and Tree Structures[C]//Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1:Long Papers).2016.DOI:10.18653/v1/P16-1105.
[4] Hang T,Feng J,Yan L,et al.Joint extraction of entities and relations using multi-label tagging and relational alignment[J].Neural Computing and Applications,2022.DOI:10.1007/s00521-021-06685-1.
[5] 赵丹丹,张俊朋,孟佳娜,等.基于预训练模型和混合神经网络的医疗实体关系抽取[J].北京大学学报:自然科学版,2023,59(1):65-75.
[6] 耿飙,梁成全,魏炜,等.基于深度学习的非结构化医学文本知识抽取[J].计算机工程与设计,2024,45(1):177-186.
[7] 夏天赐,孙媛.基于联合模型的藏文实体关系抽取方法研究[J].中文信息学报,2018,32(12):76-83.
[8] 于韬,尼玛次仁,拥措,等.基于藏文Albert预训练语言模型的图采样与聚合实体关系抽取[J].中文信息学报,2022,36(10):63-72.
[9] 周青,拥措,拉毛东只,等.基于跨度表示的藏医药文献实体关系抽取[J/OL].北京大学学报(自然科学版),1-11[2025-03-17].https://doi.org/10.13209/j.0479-8023. 2024.125.
[10] 周青,拥措,拉毛东只,等.基于预训练的藏医药实体关系抽取[J].中文信息学报,2024,38(8):76-83.
