Abstract: Semantic similarity calculation has an important role in information retrieval of linked data,and the results of calculation directly affect the effect of data mining．The attribute information of instance is an essential factor for semantic similarity computation of linked data．To solve the problem of lower computation precision caused by lack of considering the importance of attribute and type of attribute value,this paper proposed a new semantic similarity calculation method based on dynamic weight．This method dynamically computes the attribute weight according to quantity of different attribute values,distribution of attribute values,and validity of attribute．Then,it chooses the matching similarity algorithm of attributes according to the types of attribute value．Finally,it combines the dynamic weight of attributes to calculate the semantic similarity of instances．The experiment confirms that the computation precision of the semantic similarity of instances obtained from the methods in this thesis is better than existing methods.

Key words: Linked data,Semantic similarity,Instance attributes,Dynamic weight