摘要
针对传统的语义相似度计算方法计算量过大、计算过程较复杂等问题,提出了一种基于阶段递进的综合本体相似度计算方法。该方法把计算相似度的过程分为4个阶段,每个阶段根据实际情况设定一个阈值,如果此阶段计算的相似度大于阈值,则计算下一阶段的相似度;如果小于阈值,则认为该对概念间不相似,不必再计算以下各阶段的相似度,可大大减少相似度的计算量,使计算过程清晰可控。通过实验数据可知,该算法与Glue算法相比,其查全率、查准率分别提高4.78%和3.05%,而计算效率提高50%以上。
According to the semantic similarity calculation of traditional method in the presence of a large,complex calculation process problems,put forward a calculation method of comprehensive ontology similarity based on stage progression.The method to process the similarity is divided into four stages,each stage according to the actual situation of setting a threshold value,if the phase calculation of similarity is greater than a threshold,then calculate the similarity of the next phase,if less than the threshold then the concept of similarity between dissimilar,do not have to then calculate the following phases,this can greatly reduce the calculation of similarity,the calculation process and control.The algorithm in the recall and precision are increased by 4.78% and 3.05% than Glue algorithm through experiment data,while the calculation efficiency increased to 50% above.
出处
《吉林大学学报(信息科学版)》
CAS
2014年第2期201-204,共4页
Journal of Jilin University(Information Science Edition)
基金
吉林省科技厅自然科学基金资助项目(20130101060JC)
吉林省教育厅"十一五"科学技术研究基金资助项目(201046)
关键词
阶段递进
相似度
阈值
stage progression
similarity
threshold