Semantic interoperability between heterogeneous multi-agent systems based on Deep Learning

Abstract

Ontologies are important for knowledge-based information systems such as multi-agent systems. Ontologies are a natural solution to ensure a semantic interoperability between heterogeneous multi-agent systems. In this paper, we present a new model that uses a trained neural network to build ontologies adapted from other ontologies in order to solve the problem of semantic interoperability between heterogeneous multi-agent systems (SMAs). The main idea is to attribute to each concept of a given SMA ontology an image label that indicates its semantic representation. To build a new adapted ontology, a trained neural network is used to interpret the ontology concepts of an existing source SMA.