Abstract:
Stream Reasoning (SR) has emerged as a crucial paradigm for enabling real-time, intelligent decision-making over dynamic data streams, which are increasingly prevalent in domains such as IoT, edge computing, and decentralized systems. Formalisms for SR are essential because they define the theoretical and practical foundations for reasoning under continuous, time-sensitive conditions.
This paper presents a comprehensive review of SR formalisms introduced over the past 15 years, evaluating their evolution and impact. We propose a set of dimensions for assessing SR formalisms, considering both theoretical properties, such as expressivity, underlying paradigm, and stream representation, as well as practical metrics like citation count, software development, and real-world applications. Through a detailed analysis of the literature since 2009, at which the foundational paper ``It's a Streaming World'' was published, we score the existing SR formalisms based on these dimensions, highlighting the field's considerable progress.
Our findings indicate that recent advancements have led to sophisticated SR formalisms capable of tackling increasingly complex reasoning tasks.
However, further research remains essential, as no single formalism satisfies all possible requirements. The best choice depends on the specific needs of the intended application. To support this, we provide a broad overview of currently existing formalisms to help practitioners select the most suitable approach. Formalisms like DatalogMTL and LARS stand out for their strong theoretical foundations and promise for supporting advanced applications, while query-based formalisms show potential for addressing intricate reasoning challenges beyond basic querying. Additionally, SR formalisms that integrate RDF streams are particularly well-positioned to enhance interoperability across heterogeneous systems, opening concrete opportunities for deployment in IoT and edge computing scenarios.
