融合多粒度语义信息和知识图谱的中文医疗问答匹配模型

doi:10.3778/j.issn.1002-8331.2305-0453

摘要/Abstract

摘要： 中文医疗领域问答容易受到医疗特定词汇的噪声影响，相对于开放领域问答其更具有挑战性。以往的中文医疗问答研究主要依赖于字符级别的细粒度信息，忽略了携带更多语义信息的单词级别的粗粒度信息。此外，引入外部医学知识图谱可以进一步丰富问答句子中的细粒度信息，然而目前大多数研究通常只采用句子和外部知识共同表示的简单方式。由此提出一种融合多粒度语义信息和知识图谱的中文医疗问答匹配模型（CMQA-MGSI）。该模型引入Lattice网络，结合Word2Vec和BERT设计了两种特征向量提取模型来选择问答句子中最相关的字符序列和单词序列以获得更丰富的多粒度语义信息；为了更好地融合外部领域知识，设计双通道注意力模块提取问答句子和知识图谱中实体嵌入以及关系嵌入之间多个角度的知识表征信息。该模型在数据集cMedQA1.0和cMedQA2.0上的实验表明，效果优于现有的问答匹配模型。

关键词: 中文医疗问答, 多粒度信息, 知识图谱, Lattice网络, 注意力机制

Abstract: Chinese medical Q&A is easily affected by the noise of medical-specific terminology, making it more challenging than open-domain Q&A. Previous studies on Chinese medical Q&A mainly relied on character-level fine-grained information, neglecting word-level coarse-grained information that carries more semantic information. In addition, introducing external medical knowledge graph can further enrich the fine-grained information in Q&A sentences, but most existing studies usually adopt a simple way of joint representation of sentences and external knowledge. Therefore, this paper proposes a Chinese medical Q&A matching model based on multi-granularity semantic information and knowledge graph (CMQA-MGSI). The model employs a Lattice network to select the most relevant character-level and word-level sequences from the Q&A sentences, and leverages Word2Vec and BERT to enhance the semantic information; to better exploit the external domain knowledge, a dual-channel attention mechanism is devised to capture the multi-angle knowledge representations between the Q&A sentences and the entity embeddings and relation embeddings in the knowledge graph. Experi-ments on the cMedQA1.0 and cMedQA2.0 datasets demonstrate that the proposed model outperforms existing Chinese medical Q&A matching models.

Key words: Chinese medical Q&A, multi-granularity information, knowledge graph, Lattice network, attention mechanism

管立本, 李实. 融合多粒度语义信息和知识图谱的中文医疗问答匹配模型[J]. 计算机工程与应用, 2024, 60(14): 152-161.

GUAN Liben, LI Shi. Chinese Medical Q&A Matching Model Based on Multi-Granularity Semantic Information and Knowledge Graph[J]. Computer Engineering and Applications, 2024, 60(14): 152-161.

参考文献

[1] FENG M, BING X, GLASS M R, et al. Applying deep learning to answer selection: a study and an open task[C]//Proceedings of the Automatic Speech Recognition & Understanding, 2015: 813-820.
[2] CUI X, HAN J. Chinese medical question answer matching based on interactive sentence representation learning[J]. arXiv:2011.13573, 2020.
[3] ZHANG S, ZHANG X, WANG H, et al. Chinese medical question answer matching using end-to-end character-level multi-scale CNNs[J]. Applied Sciences, 2017, 7(8): 767.
[4] ZHANG S, ZHANG X, WANG H, et al. Multi-scale attentive interaction networks for Chinese medical question answer selection[J]. IEEE Access, 2018, 6: 74061-74071.
[5] CHEN X, YANG Z, LIANG N, et al. Co-attention fusion based deep neural network for Chinese medical answer selection[J]. Applied Intelligence, 2021, 51: 6633-6646.
[6] LIANG S, CHEN X, MA J, et al. An improved double channel long short-term memory model for medical text classification[J]. Journal of Healthcare Engineering, 2021: 1-8.
[7] LI S, YAO Y H. Improving medical Q&A matching by augmenting dual-channel attention with global similarity[J]. Computational Intelligence and Neuroscience, 2022, 2022.
[8] ZHANG Y, YANG J. Chinese NER using lattice LSTM[J]. arXiv:1805.02023, 2018.
[9] ZHANG Y, WANG Y, YANG J. Lattice LSTM for Chinese sentence representation[J]. IEEE Transactions on Audio Speech and Language Processing, 2020, 28: 1506-1519.
[10]YU H, LEE M, KAUFMAN D, et al. Development, implementation, and a cognitive evaluation of a definitional question answering system for physicians[J]. Journal of Biomedical Informatics, 2007, 40(3): 236-251.
[11] CAIRNS B L, NIELSEN R D, MASANZ J J, et al. The MiPACQ clinical question answering system[C]//Proceedings of the AMIA Annual Symposium, 2011: 171-180.
[12] SONAL J, TRIPTI D. Rule based architecture for medical question answering system[C]//Proceedings of the Second International Conference on Soft Computing for Problem Solving, 2012: 28-30.
[13]   HEILMAN M, SMITH N A. Tree edit models for recognizing textual entailments, paraphrases, and answers to questions[C]//Proceedings of the Human Language Technologies: Conference of the North American Chapter of the Association of Computational Linguistics, 2010: 1011-1019.
[14] LU W, WU H, JIAN P, et al. An empirical study of classifier combination based word sense disambiguation[J]. IEICE Transactions on Information and Systems, 2018, 101(1): 225-233.
[15] TOBA H, MING Z Y, ADRIANI M, et al. Discovering high quality answers in community question answering archives using a hierarchy of classifiers[J]. Information Sciences, 2014, 261: 101-115.
[16] 刘芳, 于斐. 面向医疗行业的智能问答系统研究与实现[J]. 微电子学与计算机, 2012, 29(11): 95-98.
LIU F, YU F. Research and development of intelligent Q&A system for medical industry[J]. Microelectronics & Computer, 2012, 29(11): 95-98.
[17]   SINHA S, NARAYANAN S. Model based answer selection[C]//Proceedings of the Textual Inference in Question Answering Workshop, 2005: 42-46.
[18] ATHRNIKOS S J, HAN H, BROOKS A D. A framework of a logic-based question-answering system for the medical domain (LOQAS-Med)[C]//Proceedings of the 2009 ACM Symposium on Applied Computing, 2009: 847-851.
[19] MOSCHITTI A, QUARTERONI S. Linguistic kernels for answer re-ranking in question answering systems[J]. Information Processing & Management, 2011, 47(6): 825-842.
[20]   YEN S J, WU Y C, YANG J C, et al. A support vector machine-based context-ranking model for question answering[J]. Information Sciences, 2013, 224: 77-87.
[21]   王杰. 基于深度神经网络的中文医疗社区问答匹配技术研究[D]. 成都: 电子科技大学, 2020.
WANG J. Research on question and answer matching technology of Chinese medical community based on deep neural network[D]. Chengdu: University of Electronic Science and Technology of China, 2020.
[22]   HU B, LU Z, LI H, et al. Convolutional neural network architectures for matching natural language sentences[C]//Proceedings of the 27th International Conference on Neural Information Processing Systems, 2014: 2042-2050.
[23] TAN M, SANTOS C, XIANG B, et al. LSTM-based deep learning models for non-factoid answer selection[J]. arXiv:1511.04108, 2015.
[24] WANG W, YAN M, WU C. Multi-granularity hierarchical attention fusion networks for reading comprehension and question answering[J]. arXiv:1811.11934, 2018.
[25] 张崇宇. 基于知识图谱的自动问答系统的应用研究与实现[D]. 北京: 北京邮电大学, 2019.
ZHANG C Y. Application research and implementation of automatic question answering system based on knowledge graph[D]. Beijing: Beijing University of Posts and Telecommunications, 2019.
[26] SHEN Y, DENG Y, YANG M, et al. Knowledge-aware attentive neural network for ranking question answer pairs[C]//Proceedings of the 41st International ACM SIGIR Conference on Research & Development in Information Retrieval, 2018: 901-904.
[27] DENG Y, XIE Y, LI Y, et al. Contextualized knowledge-aware attentive neural network: enhancing answer selection with knowledge[J]. ACM Transactions on Information Systems, 2021, 40(1): 1-33.
[28] 钱振飞. 融合知识图谱信息的常用医疗知识问答系统的设计与实现[D]. 镇江: 江苏大学, 2022.
QIAN Z F. Design and implementation of common medical knowledge question answering system integrating knowledge graph information[D]. Zhenjiang: Jiangsu University, 2022.
[29] LAI Y, LIU Y, FFENG Y, et al. Lattice-BERT: leveraging multi-granularity representations in Chinese pre-trained language models[J]. arXiv:2104.07204, 2021.