Retrieval performance experiment with the webspace method

Finding relevant information using search engines that index large portions of the World-Wide Web is often a frustrating task. Due to the diversity of the information available, those search engines will have to rely on techniques, developed in the field of information retrieval (IR). When focusing on more limited domains of the Internet, large collections of documents can be found, having a highly structured and multimedia character. Furthermore, it can be assumed that the content is more related. This allows more precise and advanced query formulation techniques to be used for the Web, as commonly used within a database environment. The Webspace Method focuses on such document collections, and offers an approach for modelling and searching large collections of documents, based on a conceptual schema. The main focus in this article is the evaluation of a retrieval performance experiment, carried out to examine the advances of the webspace search engine, compared to a standard search engine using a widely accepted IR model. A major improvement in retrieval performance, measured in terms of recall and precision, up to a factor two, can be achieved when searching document collections, using the Webspace Method.