Combining Data Integration and IE Techniques to Support Partially Structured Data

A class of applications exists where the information to be stored is partially structured:that is, it consists partly of some structured data sources each conforming to a schema and partly of information left as free text. While investigating the requirements for querying partially structured data, we have encountered several limitations in the currently available approaches and we describe here three new techniques which combine aspects of Information Extraction with data integration in order to better exploit the data in these applications.