Abstract: To improve the recall，precision，and efficiency of grid services discovery，based on ontology technology and rough sets theory，an algorithm OGSDA-BG is designed.The property sets of request service and advertise service act as the vertex sets of bipartite graph.The edge sets is made up of the lines between properties matched，and the matching degree is the weight of edges.So the problem of grid service discovery becomes the optimal complete match.Experiment indicates that the recall and precision of algorithm OGSDA-BG improved 10%~50%.Although the efficiency decreased about 10%，it can be accepted.

Key words: bipartite graph matching, ontology, grid services discovery, Hungarian algorithm

摘要： 为了提高网格服务发现的查全率、查准率和效率，论文设计了一个基于本体和二部图的网格服务发现算法OGSDA-BG。该算法把请求服务和发布服务的属性集分别作为二部图顶点集，所有匹配属性之间的连线为边，边权是属性匹配度，把问题转换为二部图的最优完全匹配。实验结果表明该算法的查全率和查准率较以前的算法提高了10%~50%，尽管服务发现的效率降低10%左右，但是在可接受范围之内。

关键词: 二部图匹配, 本体, 网格服务发现, 匈牙利算法