Self-Organizing Maps for Knowledge Management
(Timo Honkela, Gurusoft Oy)
Gurusoft (Espoo, Finland) creates commercial tools and solutions for information and documentation exploration, classification, search, and visualization. Solutions cover new media, e-commerce, extranets and intranets. The products are based on advanced neural network and natural language processing technology. Gurusoft technology is based on groundbreaking research on self-organizing maps and the WEBSOM method at Helsinki University of Technology.
The approach based on self-organizing maps allows detailed analysis and efficient personalisation of information flows without or with minimal human intervention. The tools even allow in-depth conceptual and dynamic analysis of any domain ranging from product areas and customer feedback to business models and decision making. All this is made available on the Web.
Gurusoft approach replaces methods that are based on, for instance, symbolic rule-based AI or specification of fixed classes or categorizations. The underlying self-organizing map applies unsupervised learning. This enables Gurusoft systems to adapt automatically and follow the conceptual changes of the application domain even with no human supervision. Moreover, the system can tune itself to the personal needs and preferences. The Gurusoft approach gives a basis for systems that are semantically autonomous and that even create a mapping between different conceptual systems. Gurusoft provides tools that make information retrieval more efficient and intuitive, tools that make handcoding of metadata unnecessary, tools that help information needs and supply to meet each other.