Merging of Topic Maps Based on Corpus

The distributed topic maps often need be merged when they are used for knowledge representation, the similarity calculation of two topics is a critical factor which affects the quality of final topic maps directly. In this paper, we present a novel approach to calculate the similarity of topics and merge the distributed topic maps, the method not only implements the syntax comparison between the topics, but constructs a domain-specific dictionary to resolve the low precision of topic semantic similarity calculation using the common dictionary purely, the massive texts are gathered form Wikipedia and Google snippets as corpus, on which the similarity score of the specific terms is calculated and stored to dictionary by a semantic text comparison method. The experiment indicates the new method can resolve particularly the problems of the common dictionary lacking many technical terms.