Semantic enrichment of product data supported by machine learning techniques

Abstract

The process of transforming big data into understandable information is the key of sustainable innovation within an Industry 4.0 factory. Machine learning techniques and cyber-physical systems are closely related to realize a new thinking of production management and factory transformation. Textual data collected in machinery logs or product documentation, does not exhibit a rich structure which can be easily understandable by both humans and machines. Therefore, data in an unstructured format needs to be enriched and transformed into a representation schema that exhibits a higher degree of structure, before it can be used and shared. The paper, introduces a novel conceptual framework to create knowledge representations from unstructured data sources, based on enriched Semantic Vectors, using a classical vector space model extended with ontological support. Hence, this research explores how traditional knowledge representations can be enriched through incorporation of implicit information derived from the complex relationships (i.e., semantic associations) modelled by domain ontologies with the addition of information presented in documents, addresses the challenges concerning data exchange and its understanding within Industry 4.0 scenarios, when supported by semantic technologies. The proposed approach is validated with industrial examples of product data used in the building and construction domain (e.g., technical specifications concerning climate control, electric power and lighting products) showing its benefits in a real-world use case.