Overcoming Fuzziness with Semantic Modeling and AI

Abstract

Today, customer service in mechanical engineering struggles with a wide variety of challenges when it comes to providing relevant information for specific use cases. The information requirements of service technicians, on the other hand, have increased enormously. On the one hand due to increasingly complex products, but also due to the growing shortage of trained specialists. Fast and efficient provision of relevant information is essential. Remote assistance systems and portal solutions are being used to meet these challenges. A heterogeneous information landscape, characterized by data silos, makes this difficult. In a preliminary study, the authors have developed an approach based on the formalization of knowledge in knowledge graphs that bridges the data silos and thus enables efficient information provision by linking information based on use cases. However, this approach requires high data quality and suitable metadata concepts. Inaccuracies in classification systems create unwanted fuzziness when linking information. The study based on this shows how the fuzziness can be eliminated by supplementary modeling of context information and by extended modeling methods in the form of so-called DynaRules. The authors also compare how AI methods and DynaRules can be used to check the plausibility of deep links with potential fuzziness.