Calculating Similarities between Tree Data Based on Structural Analysis

Abstract

In recent years, a huge amount of data is generated every day. People usually extract useful information from the data to live conveniently. In order to extract such useful information using computers, the data becomes increasingly complex such as tree and graph structures. Therefore, it is important for us to calculate similarities between the structural data for searching useful information satisfied with a user’s information need. Incidentally, the tree structure data is comparatively simple in such structured data, so that the similarity has been evaluated by simple algorithm called the tree edit distance. The algorithm is an existing method for calculating structured similarities of the tree data, and analyses the structured difference by the edit operations. However, it cannot be said that we can consider characteristics of a tree structure using this method because it uses only the edit operations. For this reason, we propose a new method for calculating similarities of the tree data considering the edit operation as well as other features; e.g. depth of a tree and the number of nodes and so on. Using our method, we can coordinate the tree edit distance and the characteristics. As a result, our method helps to calculate the similarities exactly compared with the tree edit distance algorithm.