Advanced applications of knowledge representation technology

Knowledge representation is a field of research that focuses on representing real world information in an expressive form, in order to apply reasoning tasks to the represented knowledge to solve complex problems. The Knowledge Base Paradigm states that this domain knowledge can be modelled independently of the inference tasks that are executed on this knowledge. This enables to support a variety of powerful inference tasks on the same knowledge specification.While this field of research is extensively studied in the literature, this principle rarely finds its way to practical applications in industry. In this dissertation, we investigate the practical usability of knowledge representation technologies in real life industrial applications. To this end, a number of case studies have been selected from different domains, each with different requirements. In these use cases, the practical usability of knowledge representation technologies is demonstrated. These techniques are particularly useful to provide extensive decision support to experts. The research is motivated by the many shortcomings we identified in existing approaches, whilst executing a broad variety of case studies. First, existing systems typically focus on replicating the tasks of human experts, while little attention is paid to supporting them in those tasks. Second, knowledge is seldom formulated in a readable form. However, this practice facilitates knowledge elicitation and improves transparency and maintainability of the resulting model. Third, even though knowledge representation techniques are extensively studied in the literature, they are rarely used in real world industrial applications. Throughout this dissertation, we examine how techniques from the field of knowledge representation can be used to cope with these shortcomings. According to the knowledge base paradigm, a vast variety of inference tasks can be executed on the same knowledge base. Drawing on this paradigm, we create a system that enables experts to utilize a number of customized functionalities, offering them a wide-ranging support, all based on the same domain knowledge specification. We propose the expressive, intuitive and readable cDMN notation, which enables to provide this support based on a knowledge specification that is readable and maintainable by domain experts themselves. Finally, we show that machine learning techniques can be used to analyse historical data to complement expert know-how.status: publishe