摘要
传统的案例检索方法不能反映案例之间的内在联系,导致检索结果不够准确和全面。根据食品安全案例的特点,提出一种运用知识图谱与BERT模型相结合的案例检索方法,以提高检索效果。综合考虑食品安全案例知识图谱的关系结构和实体属性特征进行案例检索。以三元组的形式表示食品安全案例并构建知识图谱,一方面,用Jaccard相似系数计算案例的关系相似度;另一方面,采用BERT模型将属性特征向量化后,计算案例属性相似度。对两部分加权求和得到案例总相似度,并进行案例检索。多组实验验证了该方法的有效性,且案例检索结果更加准确和全面。
The traditional case retrieval method cannot reflect the internal connection between cases,resulting in inaccurate and incomprehensive retrieval results.According to the characteristics of food safety cases,a case retrieval method combining knowledge graph and BERT model is proposed to improve the retrieval effect.The method comprehensively considered the relationship structure and entity attribute characteristics of food safety case knowledge graph to conduct case retrieval.Food safety cases were represented in the form of triples and knowledge graph was constructed.On the one hand,similarity coefficient was used to calculate the similarity degree of the case.On the other hand,the BERT model was adopted to calculate the similarity of the attributes of the case after the attribute features were vectored.The weighted sum of the two parts was used to obtain the total similarity of the cases,and then case retrieval was carried out.Several groups of experiments verify the effectiveness of the method,and the case retrieval effect is more accurate and comprehensive.
作者
张贤坤
李子璇
孙月
Zhang Xiankun;Li Zixuan;Sun Yue(College of Artificial Intelligence,Tianjin University of Science and Technology,Tianjin 300457,China)
出处
《计算机应用与软件》
北大核心
2021年第7期137-146,共10页
Computer Applications and Software
基金
国家自然科学基金项目(61702367,11803022)
天津市自然科学基金项目(18JCQNJC69800,19JCYBJC15300)
天津市教委科研计划项目(2017KJ033,2017KJ035,2018KJ106)。
关键词
案例检索
知识图谱
BERT模型
关系结构
权重
Case retrieval
Knowledge graph
BERT model
Relationship structure
Weight
