580 California St., Suite 400
San Francisco, CA, 94104
Key research themes
1. How can knowledge elicitation bottlenecks be addressed to enable rapid prototyping and iterative improvement of knowledge-based systems for fault diagnosis in industrial domains?
This research area focuses on developing methods to efficiently capture, formalize, and maintain domain expert knowledge for knowledge-based systems, particularly in fault diagnosis and condition monitoring within industrial applications such as power generation. The motivation arises from the significant time and expertise required in traditional knowledge elicitation, known as the knowledge elicitation bottleneck, which limits practical deployment and evolution of rule-based systems. Effective parameterisation and symbolic representation of time-series data enable not only fast initial system development but also iterative refinement as new data becomes available. This theme is critical to producing explainable, reliable diagnostic systems that can keep pace with evolving operational contexts without the drawbacks of black-box, data-driven models lacking transparency.
Key finding: This paper presents a novel approach to knowledge elicitation using symbolic primitives (rise, fall, fluctuate, stable) to parameterize time-series condition monitoring data, enabling quick and accurate formalisation of... Read more
Key finding: This empirical study highlights ongoing challenges in the practical adoption of knowledge-based systems, emphasizing that despite prior advances in KBS methods like Common-KADS, industry uptake has waned partly due to... Read more
Key finding: This work advances knowledge representation schemes capable of explicitly modeling and reasoning about procedural knowledge critical for domains like space operations where action sequences and state changes are central. By... Read more
2. What architectures and integration strategies support scalable, distributed, and multi-perspective knowledge-based systems for complex decision-making and semantic web reasoning?
This research focuses on architectural designs and integration frameworks that enable knowledge-based systems to scale efficiently over large, interconnected, or frequently changing knowledge bases. It examines networked knowledge-based systems where multiple expert modules or nodes collaborate, exploiting reusable expertise across domains, and semantic web-oriented reasoners that handle evolving ontologies and linked data. The goal is to support diverse problem-solving strategies, modular knowledge representation, and dynamic query optimization while preserving reasoning capabilities such as explanation and learning. These architectures address challenges such as sharing heterogeneous knowledge, accommodating real-time updates, and optimizing inference performance in distributed or large-scale environments.
Key finding: This foundational paper proposes a conceptual framework for constructing networks of knowledge-based systems designed to assist human decision makers in complex environments. It delineates capabilities each node must possess,... Read more
Key finding: Investigating reasoning over dynamic semantic web knowledge bases, this paper develops a backward chaining inference engine optimized for frequent knowledge base changes. The authors introduce new query optimization... Read more
Key finding: This paper presents a comprehensive architecture for knowledge base management systems (KBMS) designed to manage large, shared, and semantically rich knowledge bases supporting advanced applications. Using the Telos knowledge... Read more
Key finding: This study demonstrates a knowledge base architecture where domain and meta-level knowledge are declaratively and modularly represented to support multiple tasks including problem-solving, explanation, and learning within... Read more
3. How can knowledge-based systems incorporate active, reactive, and multi-rule paradigms to enable dynamic and context-sensitive reasoning in complex environments?
This theme explores the integration of various rule types—deductive, active, production, and event-condition-action (ECA) rules—within knowledge-based systems and databases to imbue them with proactive and responsive behaviors. Active knowledge-based systems continuously monitor for events or changes in data and autonomously trigger rules to maintain system integrity, perform automatic reasoning, or initiate control actions. The challenge lies in unifying heterogeneous rule paradigms under coherent semantics and efficient implementation strategies to support dynamic, multi-tasking applications such as monitoring, expert diagnosis, and adaptive control in real-time or evolving contexts.
Key finding: This chapter comprehensively surveys approaches to integrating multiple rule paradigms—deductive, active, and reactive rules—within knowledge base management systems to enable dynamic, event-driven behavior. It presents... Read more
Key finding: This speculative framework addresses the hierarchical nature of knowledge and reasoning required for integrated control of large scale interconnected power systems. It recognizes the limitations of purely rule-based expert... Read more