Lexical Semantic Analysis to support Ontology Maintenance Modeling of FMEA

Abstract

Listen

Translate article

For text-based documents, word representations and meaning extraction play essential roles in knowledge modeling and presentation. A maintenance procedure comprises consecutive logical arguments for determining step-by-step cause-effect events, ideally resulting in mitigative tasks. Therefore, the context and meaning representation when mimicking the purpose of a maintenance procedure are highly dependent on the word sense, syntax-semantic interface, and semantic features of the argument. This paper proposes an event-based ontology approach for supporting a failure-mode-effect analysis (FMEA), based on a lexical semantic analysis and on identifying the meaning(s) of contextual text. Our approach constructs a straightforward lexical semantic approach for analyzing the semantic and syntactic features of the contextual structures of maintenance reports, so as to facilitate translation and interpretation for knowledge-based reasoning in the format of an FMEA. Then, the knowledge is converted into a computer-understandable representation with less heterogeneity and ambiguity. The methodology enables users to obtain a representation format that maximizes shareability and accessibility, allowing for multi-purpose usage. First, it maps the argument structure into a causal event structure, in which an event is represented as a group of highly frequent contextual features or words logically linked together to shape structured arguments. Then, Dowty and Van Valin's decomposition model is employed in the format of [Event-State-Activity-Accomplishment-Result] to determine the syntax-sematic interface(s) and linking rules in the causal chain. In addition, Van Valin's model is used to differentiate between active and causative accomplishments for punctual/non-punctual changes of states in the causal chain. Finally, the metadata and/or hypernyms of causal events are represented, to accommodate ontology modeling for semantic extraction and cause-effect interpretation. We show how easily the result is converted to a computer-understandable document using Web Ontology Language & Resource Description Framework (W3C), thereby enriching interoperability and data exchange, explicit wording, multidimensional word representation, and contextual meaning extraction in machine processing.