Nstar: an interactive tool for local web search

Defining Web query languages has received much attention in the last few years. A typical system consists of a wrapper which maps a data source to a common data model and on which semantic integration is provided. However, constructing wrappers is time-consuming and can cater for tens, but not thousands of data sources. New methods should thus be investigated and then developed. In this paper, we present a novel model to retrieve information from the WWW. It is designed for ad hoc users and employs sample-based mining to extract the desirable data. Our method is more flexible than the wrapper approach.