۱- Introduction

۲- Literature review

۳- Motivation

۴- Requirements and research questions

۵- Language model

۶- Proposed architecture

۷- Taxonomy Clustering for the Semantic matching using Tree Structure algorithm (TC-SMT)

۸- Experiment and experimental results

۹- Discussion

۱۰- Conclusions

References

Abstract

The publish/subscribe model for communication has proved to be the most suitable in the Internet of things (IoT) environment because of the decoupling provided by this model that supports communication among heterogeneous parties. The standard or common publish/subscribe uses exact model to match events to subscriptions. However, in the IoT environment, an exact match is an extreme requirement because of the diverse and large environment and generation of various forms of Smart data. Therefore, semantically similar events must be considered and returned to subscribers as a possible match. However, matching events approximately to subscriptions is a much more complex task that negatively affects the efficiency of matching. Our proposed algorithm, semantic matching using the tree structure (SMT), provides efficient communication to support time-critical applications. SMT achieved linear time in terms of throughput compared with exponential time achieved in previous work. Combining SMT with taxonomy clustering improved the effectiveness in terms of the F-score, which is an indication of the recall and precision of the results, particularly when 100% of subscriptions were to be semantically matched.

Introduction

Internet of things (IoT) sensors are increasing significantly, according to Gartner forecast published in 2017 [1] ‘‘۲۰٫۴ billion connected things will be in use worldwide by 2020’’. The high rate of growth in the use of IoT in our daily lives demands effective and efficient communication between them to achieve their potential. The continuous increase in the adoption of the IoT around us in smart environments on different sized scales, such as smart homes, smart universities, and smart cities, emphasizes the need for a suitable middleware for communication. The middleware should be extendable in terms of supporting heterogeneous sensing and actuating devices. To date, there has been no such communication standard, which creates a challenge [2]. For the middleware to be usable in communication, it should support scalability, and help in crossing semantic and syntactic boundaries between communicating parties [3]. Moreover, it should be efficient given the real-time requirements of the time critical applications and the large volume of data to be communicated. According to [4], some of the most important areas of key IoT opportunities are fleet management, security, and surveillance. For example, Lufthansa Airlines is using real-time aircraft, airport, and weather sensor data to improve on-time performance and optimize operations. These areas need real-time processing because of their time-critical requirements. Moreover, most of these timecritical applications deal with big data at the same time [2]. The publish/subscribe paradigm is well suited for large-scale distributed systems because of its ability to provide scalable, efficient communication. It supports decoupling in terms of time, space and synchronization with limited resource usage [5]. Furthermore, semantic boundaries should be crossed to facilitate communication between different parties. Agreeing on a rigid unified syntax among many heterogeneous devices and users is unrealistic. Communication should be relaxed by crossing the semantic boundary and achieving approximate matching to facilitate communication between publishers and subscribers.