At present, the algorithms about semantic similarity of precise ontology are usually divided into three categories [12]: (1) Concept name based, that is, according to English morphology structure, two similar words are considered as having the same meaning. Representative algorithms are Edit-distance method and Dictionary-based method, these algorithms are simple and fast, but the environmental factor impacts on semantic information are not taken into considerations. (2) Concept instance based, that is, the joint distribution of common instances between concepts is computed using statistical probability. Representative algorithms are Glue method and Jaccard method, and these algorithms have high fitness for particular systems, but depend on the quality of training set. (3) Concept structure based, that is, based on the extracting of attributes in concept and rule relation graph and the assigning of weights, the semantic distance is calculated. Representative algorithms are Rada method and Sycara method, these algorithms are widely used, but the implicit connections of rough concepts in the real world are not involved.