Intelligent question answering knowledge subgraph matching of high-speed train component and parts knowledge graph
-
摘要: 为应对高速列车零部件知识复杂、海量且多层级的特点,提高高速列车零部件知识图谱智能问答的效果,提出了一种基于情景感知和分类模型的高速列车零部件知识图谱智能问答知识子图匹配模型。该模型通过情景模型进行情景特征提取及向量转换;再将词向量和情景向量相融合,输入到BERT(Bidirectional Encoder Representation from Transformers)模型中,进行用户问句的所属知识域分类,分类结果即为知识子图匹配的结果。经试验证明,所提模型与其他主流分类模型相比,各项性能指标更优。Abstract: In order to address the complexity, magnanimity, and multi-level characteristics of high-speed train components and parts knowledge, and improve the effectiveness of intelligent Q&A for high-speed train components and parts knowledge graph, this paper proposed a knowledge subgraph matching model of high-speed train components and parts knowledge graph based on situational awareness model and classification model. This model extracted scene features and transformed vectors through situational model, fused the word vector and situational vector, input them into the Bidirectional Encoder Representation from Transformers (BERT) model, and classified the knowledge domain to which the user question belongs. The classification results were the results of knowledge subgraph matching. Experimental results show that the proposed model has better performance indicators in all aspects compared with other mainstream classification models.
-
-
表 1 高速列车零部件实体及其所属结构词典(部分)示意
实体名称 所属结构 联轴节 转向架 空气弹簧 构架 托板 构架 表 2 训练数据(部分)
训练问句 知识域类别 今天天气如何? 非领域问句 转向架如何修? 运维域问句 转向架需求型? 需求域问句 转向架参数型? 设计参数域问句 表 3 4种模型实验结果对比
模型 宏精确率 宏召回率 宏F1值 Kg-BERT 0.97 0.97 0.96 K-BERT 0.94 0.93 0.93 BERT 0.93 0.91 0.92 BERT+情景向量 0.98 0.98 0.96 -
[1] Sheng Q Z, Benatallah B. ContextUML: a UML-based modeling language for model-driven development of context-aware Web services[C]//International Conference on Mobile Business, 11-13 July, 2005, Sydney, Australia. New York: IEEE, 2005. 206-212.
[2] Guermah H, Fissaa T, Hafiddi H, et al. A semantic approach for service adaptation in context-aware environment [J]. Procedia Computer Science, 2014(34): 587-592. DOI: 10.1016/j.procs.2014.07.077
[3] Gu T, Pung H K, Zhang D Q. A service-oriented middleware for building context-aware services [J]. Journal of Network and Computer Applications, 2005, 28(1): 1-18. DOI: 10.1016/j.jnca.2004.06.002
[4] 周维琴,石广田. 基于AutoCAD的协作虚拟感知技术的实现 [J]. 机械设计与制造,2007(10):183-185. DOI: 10.3969/j.issn.1001-3997.2007.10.078 [5] 韦斯羽,朱广丽,谈光璞. 融合Text-CNN与注意力机制的特产小吃评论情感分析 [J]. 阜阳师范大学学报(自然科学版),2023,40(1):57-63. [6] 杜文然,靳 征,代明睿,等. 基于RoBERTa-BiLSTM-CRF模型的铁路货运一口价议价策略命名实体识别 [J]. 铁路计算机应用,2023,32(5):11-15. DOI: 10.3969/j.issn.1005-8451.2023.05.03 [7] 杜 林,许传淇. 基于BERT的漏洞文本特征分类技术研究 [J]. 信息安全研究,2023,9(7):687-692. -
期刊类型引用(3)
1. 刘文斌,马子彧,张春来,徐晓磊. 基于GIS技术的铁路网能力利用分析系统设计与应用. 铁路计算机应用. 2025(03): 38-43 . 本站查看
2. 黄方圆,徐泽儒,郭禅. 基于GIS技术的多模式航运运输里程精细化分析. 中国航务周刊. 2024(48): 59-61 . 百度学术
3. 杨文澈,刘茂朕,史晓磊,董卓皇,陈亚勋. 铁路货车里程管理及辅助分析应用系统设计及应用. 铁路计算机应用. 2024(12): 59-64 . 本站查看
其他类型引用(0)