Learning query languages of Web interfaces
André Bergholz, Boris Chidlovskii · 2004
This paper studies the problem of automatic acquisition of the query languages supported by a Web information resource. We describe a system that automatically probes the search interface of a resource with a set of test queries and analyses the returned pages to recognize supported query operators. The automatic acquisition assumes the availability of the number of matches the resource returns for a submitted query. The match numbers are used to train a learning system and to generate classification rules that recognize the query operators supported by a provider and their syntactic encodings. These classification rules are employed during the automatic probing of new providers to determine query operators they support. We report on results of experiments with a set of real Web resources.