A new method based on semantic models of expert information is proposed. It aims to resolve the problem of establishing data mapping in a heterogeneous data conversion process of expert information. This paper investigates the methods of establishing a semantic model based on a data table of expert information and the basic principles and methods of semantic similarity calculation. It selects a calculation model suitable for semantics of expert data and adjusts the factors which affect the semantic similarity of the data table in this model. Through a comparative analysis of the experimental results, a more effective and accurate method of calculating the semantic similarity is found.