OntoGather: Intelligent search in dynamic data streams

This was (and still is) well-suited for static Web pages that remain stable for a longer period of time, but today many resources on the Web are in constant flux. Prominent examples for this are virtual market places or real-time information systems, e.g. stock-exchange price services. The underlying volatile nature of the aforementioned domains necessitates a dynamic approach to support user queries of the form what is the cheapest price for an IXUS digital camera (at the moment)?

To solve this problem we propose an approach that relies on dynamic integration of information sources which are accessed at query time. The core of our system is Florid [1], a deductive object-oriented database system based on F-Logic [2], which operates on top of a domain-specific background ontology. It serves as inference engine used for Web resource selection and evaluates a user query on up-to-the-minute information. In this project we focus on information extracted from HTML pages with the aid of a wrapper tool. Web resources that can be accessed via Web Service interfaces, such as WSDL APIs or RSS feeds have not been considered yet, but can be easily integrated into our system. Wrapper tools range from semi-automatic approaches,such as LIXTO [3], to fully-automatic.

In our scenario a fully-automatic approach is most suitable, because the wrapper generation and maintenance effort is negligible, which suits the requirements of our dynamic world scenario best. Therefore we use our fully-automatic extraction system ViPER [4], that is able to extract up-to-date information from arbitrary HTML pages consisting of data records, which have a similar structure.

http://dbis.informatik.uni-freiburg.de/index.php?project=OntoGather