mineXML

As industrial strength XML databases emerging, the amount of available XML information becomes tremendous. An interactive system that can automatically analyze XML data, and find meaningful and useful information within particular application domains, is needed.

Goal

The project mineXML is designed to extract high-level information from the Web Service description (XML documents) repositories for purpose of establishing associations, building classifications, and extracting meaningful information of various types (i.e., business related, configuration related).

Following are some techniques and issues we are investigating:

• XML Schema Integration -- When the XQuery schema is different from the schema of the XML documents it querying, or the XML documents have different schemas, some automatic action needs to be taken to migrate the schemas or decide that the schemas are not migratable
  Example: how do we know when different seeming items in different schemas may actually correspond, such as "state" in the US and "province" in Canada?
  
  
• Mining
  
  
  • Association -- Find interesting association or correlation relationships among a large set of data items; the rules so mined can be used for cross-market analysis, correlation analysis, and so on
    Example: what services customers are likely to use together with service A?
    
    
  • Classification -- Extract models describing important data classes and predict categorical labels; the model so constructed is used to classify future data and develop a better understanding of data
    Example: what services customers with characteristic A tend to use?
    
    
• Tooling -- Visualize, and navigate through, those extracted views of the services.

Collaboration

This project is a joint effort by Cindy Chen, Judah Diament, George Mihaila, Haixun Wang and Isabelle Rouvellou, at IBM T. J. Watson Research Center, Hawthorne, New York, USA.

For further information on mineXML, please contact Cindy Chen.