Abstract:

The present paper presents a new method of information extraction from the Deep Web based on model matching. It extracts the characteristic vector of the Deep Web query interface by means of analysising the depth of feature of web page structure automatically. The frequency and concentration rate are both considered when the weight in vector space model is defined. The characteristic word vector is used to construct the database query interface with the number of characteristic word taken into account. At last, model matching is used to classify different databases. This method is validated by experiment results.

Key words: Deep Web, data integration, model matching