Thematic event processing

Abstract

Event-based systems follow a decoupled mode of interaction between event producers and consumers in space, time, and synchronization to enable scalability within distributed systems. We recognize a fourth dimension of coupling due to the need for mutual agreements on terms that describe event types, attributes, and values. Semantic coupling is challenging in large-scale, open, and heterogeneous environments such as the Internet of Things (IoT). It requires event producers and consumers to agree on event semantics and can limit scalability due to the difficulties in establishing such agreements. In this paper we propose a new thematic event processing approach based on enhancing events and subscriptions with terms representing their themes to clarify their domains and meanings in addition to their payload. Experiments conducted using large heterogeneous sets of smart-city and energy management events suggest up to 85% of matching accuracy at a rate of 500 events/sec of throughput. This represents around 15% improvement in accuracy and 150% in throughput over non-thematic approaches. This suggests the viability of thematic event processing to scale to environments such as the IoT.

Big Data Acquisition

“…Works on semantic event processing such as semantic approximations (Hasan and Curry 2014a), thematic event processing (Hasan and Curry 2014b), and thingsonomy tagging (Hasan and Curry 2015) are emerging approaches in this area, within this context. • In order to perform post-and pre-processing of acquired data, the current state-of the art provides a set of open-source and commercial tools and frameworks.…”

Tackling variety in event-based systems

“…In [22] we proposed an approach for loose semantic coupling based on the use of distributional semantics, free tagging, and approximation. The thematic event processing approach suggests associating thematic tags that describe the themes of types, attributes, and values to clarify their meanings as shown in Figure 3.…”

“…For example, in [22] we have synthesized a set of around 15, 000 events of up to 10 attribute-value pair per event, and around 100 approximate subscriptions from real world smart city deployments in Europe such as the SmartSantander project [38] which employs a set of sensors to monitor temperature, noise, traffic, parking, etc. Figure 2 illustrates the resulting effectiveness and efficiency of the approximate matcher working with a Wikipediabased esa.…”

“…The 100 approximate subscriptions would cost users an equivalence of around 48, 000 exact subscription rules. More details on the approach can be found in [22].…”

“…Evaluating IoT semantic normalization: effectiveness (left) and time efficiency (right)[22]. Mode: free use of well representative themes Loose Coupling Mode: lightweight agreement on themes…”