OntoSpect: IoT Ontology Inspection by Concept Extraction and Natural Language Generation
Proceedings of the 21th International Conference on Web Engineering
One of the main challenges in the Internet of Things (IoT) is the lack of semantic interoperability between heterogeneous sources. In the Semantic Web domain, ontologies are one way to achieve semantic interoperability by using a common vocabulary that represents heterogeneous sources. However, recent studies have shown that the amount of concept reuse from existing IoT ontologies is low. As the number of IoT ontologies increases, encouraging users to reuse existing ontologies instead of creating new concepts becomes important. Ontology catalogues are a prominent approach to discover and inspect existing ontologies for reuse. However, such catalogues inspect the ontologies using general criteria which is not enough to understand the content of the ontology. In this paper, we propose a method for automatic ontology inspection (OntoSpect) of IoT ontologies from different application domains based on a generic set of content-related concepts. OntoSpect consists of two main steps: first it extracts the set of IoT concepts, and then generates human-understandable descriptions using a Model-driven Engineering (MDE) approach. We evaluate the quality of concept extraction and natural language description generation with 84 ontologies retrieved from the LOV4IoT catalogue and report on quality metrics. In addition, we conduct an empirical study with 28 ontology users to further assess the quality of the generated descriptions. The results demonstrate the capability of OntoSpect to support ontology users inspecting IoT ontologies.
Noura, Mahda; Wang, Yichen; Heil, Sebastian; Gaedke, Martin: OntoSpect: IoT Ontology Inspection by Concept Extraction and Natural Language Generation. Proceedings of the 21th International Conference on Web Engineering, 2021.