基于本体-匹配双向扩展模型的设计知识推荐方法  

Design knowledge recommendation method based on ontology-matching bi-directional extension model

作　　者：董锦锦 顾海瑞 陆佳炜 王琪冰 姜雪明 肖刚 DONG Jinjin;GU Hairui;LU Jiawei;WANG Qibing;JIANG Xueming;XIAO Gang(College of Mechanical and Electrical Engineering,China Jiliang University,Hangzhou 310018,China;Zhejiang Supcon Information Co.,Ltd.,Hangzhou 310018,China)

机构地区：[1]中国计量大学机电工程学院,浙江杭州310018 [2]浙江中控信息产业股份有限公司,浙江杭州310018

出　　处：《机电工程》2024年第10期1793-1805,共13页Journal of Mechanical & Electrical Engineering

基　　金：国家自然科学基金资助项目(61976193);浙江省“尖兵”“领雁”研发(攻关)计划项目(2023C01022)。

摘　　要：针对自动扶梯桁架设计知识缺乏有效归纳与整理,知识重用性与共享性差等问题,提出了一种基于本体-匹配双向扩展模型的自动扶梯桁架设计知识推荐方法。首先,提出了自动扶梯桁架命名实体识别(ETNER)模型,进行了该领域知识特征的有效提取,为知识本体构建提供了数据基础;然后,构建了领域知识本体,对设计知识进行了规范表示,并利用本体知识之间的关联关系进行了推理;最后,设计了本体-匹配双向扩展模型,在本体推理为空的场景下,采用改进知识相似度计算与文档扩展匹配的方法,提高了对于用户知识需求的理解能力,补充了推荐结果,同时促进了自动扶梯桁架设计知识本体的更新。研究结果表明:实体识别模型提取扶梯桁架设计知识特征的F1指标达到0.863,能有效提取该领域设计知识特征;该知识推荐方法的平均倒数排名(MRR)、命中率(HR)、归一化折损累计增益(NDCG)指标分别为0.79、0.85、0.80,与传统Doc2query方法相比,其可以充分理解设计知识需求,提高了知识推荐结果的有效性;该研究结果可以为扶梯桁架设计与其他类似工程应用提供知识共享及知识重用的方法。To address the lack of effective summarization and organization of knowledge on escalator truss design,which led to poor reusability and sharing of knowledge,an ontology-matching bi-directional extension model based on escalator truss design knowledge recommendation method was proposed.First,the escalator truss name entity recognition(ETNER)model was proposed,and the knowledge features in this domain were extracted effectively,which provided the data basis for the construction of knowledge ontology.Then,the domain knowledge ontology was constructed,the design knowledge was represented normally,and the correlation between the ontology knowledge was used for reasoning.Finally,the ontology-matching bidirectional extended knowledge recommendation model was designed,and in the scenario where the ontology reasoning was empty,the improved knowledge similarity calculation method and document extension matching method were used to improve the understanding of the users knowledge requirements and to supplement the knowledge recommendation results of escalator truss design,and at the same time promote the updating of the knowledge ontology of escalator truss design.The research results show that the F1 index of the proposed entity recognition model can reach 0.863,which can effectively extract the design knowledge features of the escalator truss.The mean reciprocal rank(MRR),hit ratio(HR)and normalized discounted cumulative gain(NDCG)indexes of the proposed knowledge recommendation method are 0.79,0.85 and 0.80,respectively.Comparing with the traditional Doc2query method,it can fully understand the design knowledge requirements and improve the effectiveness of knowledge recommendation results.The research results can provide a method support for knowledge sharing and reuse in escalator truss design and other similar engineering applications.

关 键 词：扶梯桁架设计 自动扶梯桁架名称实体识别模型 知识推荐 知识重用性 知识共享性 本体-匹配双向扩展模型 

分 类 号：TH236[机械工程—机械制造及自动化] TH186