Abstract: Information is mainly represented in tabular form in the modern society and scientific filed, so more and more attentions have been paid on how to extract ontology instances from the Web tables. In order to improve the accuracy of extracting ontology instances from the Web tables, an acquisition method of semantic class of words was proposed on the basis of semantic similarity. The method is based on SVM semantic similarity calculation method, which improves the accuracy of checking the semantic similarity, and overcomes the limitations of the previous analysis method of table structure by using syntactic similarity. Finally, according to the experimental results, the performance of this method is evaluated. Experimental results show that this method can effectively extract the ontology instances from Web tables.