Using linked open data to bootstrap corporate knowledge management in the OrganiK project

Gunnar Aastrand Grimnes, Remzi Çelebi, Leo Sauermann

In: Adrian Paschke, Nicola Henze, Tassilo Pellegrini (Hrsg.). Proceedings of the 6th International Conference on Semantic Systems. International Conference on Semantic Systems (I-Semantics-10) September 1-3 Graz Austria Seiten 18-1 ACM International Conference Proceeding Series ISBN 978-1-4503-0014-8 ACM 2010.


Tagging has become a wide-spread tool for organising content, from photos and music, to research paper and data-visualisations. Organising tags in a taxonomy adds hierarchical structure and relationships, this can be helpful, both for finding and applying tags to new content, as well as for enabling when searching. However, taxonomies can be very time-consuming to create and maintain. If a hierarchical taxonomy could be automatically built and adapted to a particular domain, the entry cost for using taxonomies for structuring information would go down. Small and medium enterprises (SMEs) do not currently have sufficient resources to invest in Enterprise 2.0 technologies like taxonomies, wikis or blogging as the entry cost it too high. The organik project aims to make Enterprise 2.0 features available with low entry- and maintenance costs. In this paper, an algorithm and methodology to automatically create and maintain taxonomies is presented. It analyses enterprise document corpora and uses background information from domain-specific data sources or from the Linked Open Data cloud to improve and contextualise the created SKOS taxonomy. Content created in a Drupal-based Enterprise 2.0 content management system is automatically categorised, and the automatically created taxonomy is extended when needed. The system has been tested with corpora of medical abstracts, computer science papers, and the Enron email collection, and is in productive use.

Weitere Links

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence