Estimation of Structural Similarity of XML Document Based on Frequency and Path

With the continuous development of Internet and rich resources emerging on the Web, information retrieval based on XML has emerged; the similarity of documents is the basis of information retrieval. A method is proposed to compute similarity of XML documents based on path and frequency in the paper. XML document is expressed as a collection of tuple, the paths are extracted and delete the recurring in order to improve efficiency, tag is matched by WordNet; and then path similarity is computed by the fuzzy longest common subsequence and frequency; finally, the structure similarity between documents are calculated. Two experiments are done to show that the method is effective, the experiment 1 test structural similarity of 15 XML documents from 3 DTDs; the similarity computing is applied in the documents classification for real data sets in the experiment 2, and results show the accuracy may arrive at 100%.