Agent-Based Semantic Search at motoso.de

摘要

Searching for information in large rather unstructured real-world data sets is a difficult task, because the user expects immediate responses as well as high-quality search results. Today, existing search engines, like Google, apply a keyword-based search, which is handled by indexed-based lookup and subsequent ranking algorithms. This kind of search is able to deliver many search results in a short time, but fails to guarantee that only relevant data is presented. The main reason for the low search precision is the lack of understanding of the system for the original user intention of the search. In the system presented in this paper, the search problem is tackled within a closed domain, which allows semantic technologies to be used. Concretely, a multi-agent system architecture is presented, which is capable of interpreting a key-words based search for the car component domain. Based on domain specific ontologies the search is analyzed and directed towards the interpreted intentions. Consequently, the search precision is increased leading to a substantial improvement of the user search experience. The system is currently in beta state and it is planned to roll out the functionality in near future at the car component online market-place motoso.de.