In recent years, XML has been increasingly become an emerging standard and widely used in many applications. For example, office documents which are more and more popular used at this time, are also stored in multiple parts of XML archive formats. It is known that the structure and content of XML files play different roles depending on kind of documents. Therefore, achievement similarity search of an XML file should base on both structure and content. In previous work, LAX+ is an algorithm for reckoning a similarity value from structure and contents of XML files in the office documents. However, since LAX+ used exactly matching method between corresponding leaves, similar words in the leaf-nodes are considered as different. To solve the problem, we propose to combine LAX+ with keyword similarity in leaf-nodes. We use docx, xlsx and pptx file formats as experimental data set. The evaluation shows that our approach can be used to improve the precision and recall.
各種検索
サポート
ヘルプ