应用于电网资产模型搜索的交叉权积文本相似度计算方法  

A cross weight product text similarity calculation method for power grid model search

作　　者：赵增涛 张豪 余益龙 ZHAO Zengtao;ZHANG Hao;YU Yilong(CSG Power Generation Company,Guangzhou 510630,Guangdong,China;Weihai CIMS Tech Co.,Ltd,Weihai 264209,Shandong,China)

机构地区：[1]南方电网调峰调频发电有限公司,广东广州510630 [2]威海欣智信息科技有限公司,山东威海264209

出　　处：《水利水电技术（中英文）》2020年第S02期209-214,共6页Water Resources and Hydropower Engineering

基　　金：南方电网科技项目(020000MS62190005)

摘　　要：电网资产模型搜索中对搜索结果的排序,采用了按综合得分由高到低排列的方法。综合得分由多字段综合文本相似度得分、过滤条件匹配度得分、关注热度得分,按照一定的占比计算得到。多字段综合文本相似度算法的基础是短文本的相似度计算方法,需要根据电网资产模型中各个不同领域数据的特点进行灵活调整。设计出具有一定可调节性的短文本相似性计算方法。算法构建两个与需要计算相似度的两个短文本字符长度相同的权重数组并赋予初识权重值,再遍历其中一个字符串中的字符,根据字符是否在另外一个字符串中是否存在调整其权重值,之后对单字匹配、连续匹配的字符计算权重交叉乘积获得文本相似性权重,与原始权重积相除获得文本相似度值。应用交叉权积相似性算法的电网资产模型搜索,在搜索结果的准确性方面更贴近电力专业用户的期望。The ranking of the search results in the grid model search uses the method of ranking by comprehensive score from high to low.The comprehensive score is calculated by multi-field comprehensive text similarity score,filter matching score,and attention score,calculated according to a certain percentage.The basis of the multi-field comprehensive text similarity algorithm is the similarity calculation method of short text,which needs to be flexibly adjusted according to the characteristics of various fields of data in the grid model.Therefore,a short text similarity calculation method with certain adjustability is designed.The algorithm constructs two weight arrays with the same length as the two short text that need to be calculated for similarity and assigns the initial weight value,then traverses the characters in one of the strings,and adjusts their weight according to whether they exist in another,then calculates the weight cross product of single word matching and continuous matching characters to obtain the text similarity weight,and obtains the text similarity value by dividing the product of the two original text total weights.The grid model search based on cross similarity algorithm is closer to the expectation of power system users in terms of the accuracy of search results.

关 键 词：电网资产模型 搜索 文本相似性 文本权重 交叉权积 

分 类 号：F426.61[经济管理—产业经济] TP391.1[自动化与计算机技术—计算机应用技术]