XML is the new standard for information exchange and retrieval. An XML document has a schema that defines the data definition and structure of the XML document (Abiteboul et al., 2000). Due to the wide acceptance of XML, a number of techniques are required to retrieve and analyze the vast number of XML documents. Automatic deduction of the structure of XML documents for storing semi-structured data has been an active subject among researchers (Abiteboul et al., 2000; Green et al., 2002). A number of query languages for retrieving data from various XML data sources also has been developed (Abiteboul et al., 2000; W3c, 2004). The use of these query languages is limited (e.g., limited types of inputs and outputs, and users of these languages should know exactly what kinds of information are to be accessed). Data mining, on the other hand, allows the user to search out unknown facts, the information hidden behind the data. It also enables users to pose more complex queries (Dunham, 2003). Figure 1 illustrates the idea of integrating data mining algorithms with XML documents to achieve knowledge discovery. For example, after identifying similarities among various XML documents, a mining technique can analyze links between tags occurring together within the documents. This may prove useful in the analysis of e-commerce Web documents recommending personalization of Web pages.
TopIntroduction
XML is the new standard for information exchange and retrieval. An XML document has a schema that defines the data definition and structure of the XML document (Abiteboul et al., 2000). Due to the wide acceptance of XML, a number of techniques are required to retrieve and analyze the vast number of XML documents. Automatic deduction of the structure of XML documents for storing semi-structured data has been an active subject among researchers (Abiteboul et al., 2000; Green et al., 2002). A number of query languages for retrieving data from various XML data sources also has been developed (Abiteboul et al., 2000; W3c, 2004). The use of these query languages is limited (e.g., limited types of inputs and outputs, and users of these languages should know exactly what kinds of information are to be accessed). Data mining, on the other hand, allows the user to search out unknown facts, the information hidden behind the data. It also enables users to pose more complex queries (Dunham, 2003).
Figure 1 illustrates the idea of integrating data mining algorithms with XML documents to achieve knowledge discovery. For example, after identifying similarities among various XML documents, a mining technique can analyze links between tags occurring together within the documents. This may prove useful in the analysis of e-commerce Web documents recommending personalization of Web pages.
TopBackground: What Is Xml Mining?
XML mining includes mining of structures as well as contents from XML documents, depicted in Figure 2 (Nayak et al., 2002). Element tags and their nesting therein dictate the structure of an XML document (Abiteboul et al., 2000). For example, the textual structure enclosed by <author>… </author> is used to describe the author tuple and its corresponding text in the document. Since XML provides a mechanism for tagging names with data, knowledge discovery on the semantics of the documents becomes easier for improving document retrieval on the Web. Mining of XML structure is essentially mining of schema including intrastructure mining, and interstructure mining.
Intrastructure Mining
Concerned with the structure within an XML document. Knowledge is discovered about the internal structure of XML documents in this type of mining. The following mining tasks can be applied.
The classification task of data mining maps a new XML document to a predefined class of documents. A schema is interpreted as a description of a class of XML documents. The classification procedure takes a collection of schemas as a training set and classifies new XML documents according to this training set.
The clustering task of data mining identifies similarities among various XML documents. A clustering algorithm takes a collection of schemas to group them together on the basis of self-similarity. These similarities are then used to generate new schema. As a generalization, the new schema is a superclass to the training set of schemas. This generated set of clustered schemas can now be used in classifying new schemas. The superclass schema also can be used in integration of heterogeneous XML documents for each application domain. This allows users to find, collect, filter, and manage information sources more effectively on the Internet.
The association data mining describes relationships between tags that tend to occur together in XML documents that can be useful in the future. By transforming the tree structure of XML into a pseudo-transaction, it becomes possible to generate rules of the form “if an XML document contains a <craft> tag, then 80% of the time it also will contain a <licence> tag.” Such a rule then may be applied in determining the appropriate interpretation for homographic tags.