A deep relation learning method for IoT interoperability enhancement within semantic formalization framework

Abstract

Internet of Things (IoT) is facing with the interoperability issue due to the massive amount of heterogeneous entities (both physical and virtual entities) constantly generating heterogeneous data objects; semantic formalization has been widely recognized as a basis for the IoT interoperability, by which IoT can acquire the ability to comprehend data and further recognize the logic relations among heterogeneous IoT entities and heterogeneous data objects, thus to establish mutual understanding between each other to support with interoperability. Even semantic-driven track has emphasizes a lot on the logic relations in connection to the service rules and policies for interoperability, it is important that the quantity-driven relations should be also explored with adhering to the framework of semantic formalization. This paper explores a Deep Recursive Auto-encoders formed data relation learner in line with the semantic framework, which supports the data interoperability enhancement in a quantity-driven way based on the logic-driven framework. The learner starts with representing the virtual IoT entities via feature extraction; based on that, learner is trained in a manner of considering the surrounding relations of the targeted entity. As a baseline, a contrast learner with "regular" structure has been proposed which cannot functionally support semantic framework, even though the semantic formalization is indispensable; regardless the limitations in lab environment, the conducted experiments show that the proposed learner performs a bit better than the contrast learner under the same conditions. So that, the proposed method can synergistically enhances the interoperability within a semantic formalization framework.