一种基于最大权匹配计算的信息检索方法

基于《知网》的词汇语义计算方法,提出了一种基于向量空间模型的文本信息检索新方法。方法的基本技术思想是通过计算关键词的语义相似度,并采用最大权匹配方法来计算查询向量和文本向量的相似度,作为相关文本的检索依据。该方法基于全局最优,使文本和查询向量中各词条的相似度总和最大,从而可以从整体上提高文本信息检索的准确率。论文还通过原型实验对该方法的有效性进行了验证。

Method of information retrieval based on computing maximum-weight-matching

Based on the computation of words-semantic similarity of "How-net" ,a new method of document information retrieving based on the model of vector space has been proposed.The basic ideas of the method are that,firstly we compute the similarity between keywords by the words-semantic similarity of "How-net" ,and secondly compute the similarity between the query vector and the document vector based on computing the maximum-weight-matching.This algorithm is based on the optimization on overall situation and it might find the maximum sum of each pair terms' similarity between the query vector and the document vector.The experiment result has demonstrated the validity of the algorithm.