Automatic construction of a semantic, domain-independent knowledge base

Abstract

In this paper, we want to show which difficulties arise when automatically constructing a domain-independent knowledge base from the web. We show possible applications for such a knowledge base to emphasize its importance. Current knowledge bases often use manually-built patterns for extraction and quality assurance which does not scale well. Our contribution to the community will be a technique to automatically assess extracted information to ensure high quality of the information and a method of how the knowledge base can be kept up to date. The research builds upon the existing WebKnox system for Web Knowledge Extraction which is able to extract named entities and facts from the web. This is a position paper. © Springer-Verlag 2009.