Semantics-based exploratory search on manufacturing knowledge

Abstract

Knowledge graphs have been identified within the manufacturing industry as a suitable technology for integrating multidisciplinary knowledge from heterogeneous data sources. Effectively reusing this knowledge can enhance decision-making processes, and thereby fos- tering a competitive edge. However, there has been limited attention given in production research to the active human involvement in exploring these manufacturing knowledge graphs. In this context, exploratory search systems present a promising solution to encourage such engagement. Nevertheless, the majority of exploratory search systems are designed for general knowl- edge graphs where common knowledge suffices. Within the multifaceted and complex landscape of manufacturing, it is crucial to focus more closely on specific exploratory search features. Drawing from interviews with domain experts, this thesis highlights three distinct features that boost the efficacy of exploratory search for manufacturing environments. These include the system’s ability to tailor itself to numerous engineering perspectives, the availability of transparent provenance information that notably assists engineers in their investigative work, and the need for comprehending as well as navigating complex and deep hierarchical structures. An exploratory search system (ESS) has been designed, informed by an examination of common interface paradigms for generic exploratory search systems, coupled with an exploratory analysis of paradigms equipped to meet the identified requirements for the manufacturing domain. This system encompasses a rendering engine that supports an adaptive user interface, enabling a multifaceted configuration of visualizations and the underlying search algorithms. To address the need for provenance visibility, the system is structured to disseminate named graph information pertinent to each statement to the relevant UI components. A simple tree view was employed to overcome the challenge of navigating through hierarchical structures. This ESS prototype was evaluated on a small scale with five participants for the pilot factory use case. The initial assessment of the ESS demonstrated its value for users executing exploratory search tasks in the domain of collaborative robotics. Potential areas for enhancement were identified, however. Users perceived the tree view design as not useful, and they frequently opted for managing multiple browser tabs over utilizing the memorization feature. Despite these potential areas for improvement, users were overall successful in achieving their search objectives.