Scalable integration of uncertainty reasoning and semantic web technologies

In recent years formal logical standards for knowledge representation to model real world knowledge and domains and make them accessible for computers gained a lot of trac- tion. They provide an expressive logical framework for modeling, consistency checking, reasoning, and query answering, and have proven to be versatile methods to capture knowledge of various fields. Those formalisms and methods focus on specifying knowl- edge as precisely as possible. At the same time, many applications in particular on the Semantic Web have to deal with uncertainty in their data; and handling uncertain knowledge is crucial in many real- world domains. However, regular logic is unable to capture the real-world properly due to its inherent complexity and uncertainty, all the while handling uncertain or incomplete information is getting more and more important in applications like expert system, data integration or information extraction. The overall objective of this dissertation is to identify scenarios and datasets where methods that incorporate their inherent uncertainty improve results, and investigate approaches and tools that are suitable for the respective task. In summary, this work is set out to tackle the following objectives: 1. debugging uncertain knowledge bases in order to generate consistent knowledge graphs to make them accessible for logical reasoning, 2. combining probabilistic query answering and logical reasoning which in turn uses these consistent knowledge graphs to answer user queries, and 3. employing the aforementioned techniques to the problem of risk management in IT infrastructures, as a concrete real-world application. We show that in all those scenarios, users can benefit from incorporating uncertainty in the knowledge base. Furthermore, we conduct experiments that demonstrate the real- world scalability of the demonstrated approaches. Overall, we argue that integrating uncertainty and logical reasoning, despite being theoretically intractable, is feasible in real-world application and warrants further research