Searching for Knowledge Instead of Web Sites

Today’s Web search engines can find Web pages that contain certain keywords. Up to now, however, any advanced information demands that concern facts from multiple Web pages, let alone a logical connection between them, are inherently beyond the answering capabilities of search engines. This is why our approach is to collect information from Web sites and to organize it in a huge knowledge structure, an ontology. Using pattern extraction, structural analysis and statistical learning methods, we have developed tools that can automatically build and maintain such an ontology from the contents of the Wikipedia encyclopedia and other Internet sources. Our ontology, coined Yago, contains about one million entities and concepts, and knows more than six million facts about them. Yago also has a Web interface for answering knowledge queries online. Yago could be the starting point for a new generation of search engines, for searching the Web as well as digital libraries and e-science repositories. This article describes the state of Yago in the year 2007. It has been translated from German by Krista Ames. Searching on the Web The internet has developed into a significant source of information over the last decade. Train schedules, news, academic articles, company data, and even entire encyclopedias are available online. The majority of these web pages is indexed by search engines. For example, Google enables us to search for keywords on billions of internet pages in a few split seconds. This technology is entirely sufficient for a lot of enquiries. We normally find what we are looking for after browsing through the search results for a short while. If, for example, you search for the physicist Max Planck, Google gives you a link to the Max Planck Society and several of the physicist’s biographies. Even questions like