ABSTRACT. Biology represents a very challenging domain that is typically tackled by experts in the field, with few or no interactions with the Web knowledge and rules interoperation community. However, there has been a considerable growth of data regarding biological aspects in the last decades. Moreover, the COVID-19 pandemic has traced an unprecedented point in history, where tons of information have been collected in laboratories worldwide and deposited into open data banks. Inspired by the current needs and backed by a solid knowledge base (our extensional knowledge source) called CoV2K, we propose to express and resolve a series of problems related to the SARS-CoV-2 virus and its interpretation. We formulate our queries as rules in Vadalog (our knowledge representation and reasoning language) and input them to its related logic-based reasoning system. Four cases are presented that allow to explore 1) variants effects and how they are explained in scientific literature; 2) the most typical mutations of a variant; 3) the most likely acquisition of a new mutation by a given variant and the associated reported effects; 4) the most relevant mutations of the virus according to the community. Expressing biological problems using a logic formalism is a major challenge, due to the intrinsic complexity of the domain. The four use cases show that a logical formalism is effective in expressing relevant problems for understanding the current evolution of SARS-CoV-2 variants, an essential aspect of the COVID-19 pandemic.

ABSTRACT. We present a GDPR-compliant data privacy and access use case of a distributed data wallet and we explore its modeling using two options, AspectOWL and Prova. This use case requires a representation capable of expressing the dynamicity and interaction between parties. While both approaches provide the expressiveness of non-monotonic states and fluent state transitions, their scope and semantics are vastly different. AspectOWL is a monotonic context ontology language, able to represent dynamic state transitions and knowledge retention by wrapping parts of the ontology in isolated contexts, while Prova can handle state transitions at runtime using non-monotonic state transition semantics. We present the two implementations and we discuss the similarities, advantages, and differences of the two approaches.

ABSTRACT. Rules and world knowledge are important priors in decision making in our daily lives and similarly, autonomous driving (AD) systems can exploit knowledge in their driving decision making and conflict resolution. In this doctoral thesis, the goal is to investigate ways to represent and formalize the traffic rules and regulation which exists in natural language text to integrate them with machine learning based trajectory prediction module for conflict resolution among traffic participants. So that the long-term trajectory (3 to 5 Seconds horizon) of the vehicle can be predicted using hybrid learning, which leverages both rules and data into the ML models. As only rule-based systems are not able to represent complex interactions among various traffic scene participants. On the other hand, learning-based approaches manage to represent complex interaction but they need a lot of data and in many cases, it is not feasible to generate corner case data. Recent research in combining rules with data (Neuro-Symbolic model) is an exciting research direction to leverage the best of both worlds for reasoning over traffic scene. Our work will investigate the impact of the world and commonsense knowledge integration into ML models for a safer autonomous driving experience. Apart from this, we will be working toward finding optimal representation of traffic rules for hybrid learning tasks and leverage large language models for representation of traffic rules and regulation for automated rules representation for a downstream task like AD. Such kind of approach will reduce dependency on data, generate traffic rule complaint trajectory of the vehicle, and makes the model more generalizable even for corner cases or less representative cases in datasets due to reasoning capability.

ABSTRACT. It is crucial to validate driving behaviour against current road rules to improve Autonomous Vehicle (AV) safety. Due to the quantitative expressions of Queensland Overtaking Road Rules, including vagueness, exceptions, and potential rule conflicts, it is challenging to validate AV behaviour. This research introduces a Defeasible Deontic Logic (DDL) based validation system to validate AV driving behaviour against current overtaking road rules. The overtaking maneuver is the case study to illustrate the usefulness of the proposed validation system. The evaluation shows the effectiveness of the proposed validation system.

ABSTRACT. Knowledge Graphs have rapidly become a mainstream technology that combines features of databases and AI. In this talk I will introduce Knowledge Graphs, explaining their features and the theory behind them. I will then consider some of the challenges inherent in both the theory and implementation of Knowledge Graphs and present some solutions that have made possible the development of popular language standards and robust and high-performance Knowledge Graph systems. Finally, I will illustrate the wide applicability of knowledge graph technology with example use cases including configuration management, fraud detection, semantic search & browse, and data wrangling.

ABSTRACT. Warded Datalog+/- is a powerful member of the Datalog+/- family, which extends the logic language Datalog with existential quantification and provides full support for recursion. Such expressive power, paired with a promising trade-off with the offered data complexity, was the catalyst for the recent rise of the language as a relevant candidate for knowledge graph traversal and ontological reasoning applications. Despite the growing research and industrial interest towards Warded Datalog+/-, we observe a substantial lack of specific tools able to generate non-trivial settings and benchmark scenarios, essential to evaluate, analyze and compare reasoning systems over such tasks. In this paper, we aim at filling this gap by introducing iWarded, a versatile generator of Warded Datalog+/- benchmarks. Our system is able to efficiently create very large, complex and realistic reasoning settings while providing extensive control over the theoretical underpinnings of the language. iWarded was developed and employed in the context of the Vadalog system, a state-of-the-art Warded Datalog+/- -based reasoner.

ABSTRACT. The need for reasoning over temporal data has recently emerged. DatalogMTL is a highly suitable language to handle many real-world applications. In spite of the deep theoretical contribution and the first experimental implementations of DatalogMTL, practical temporal reasoning applications call for a fully engineered system, able to reason with DatalogMTL while supporting a number of features of fundamental utility such as recursion, aggregation, and negation. We introduce Temporal Vadalog, a new reasoning system for DatalogMTL that is capable of handling, among other elements, stratified negation and a form of aggregation. We evaluate the system in real-world and synthetic scenarios, comparatively showing its performance.

ABSTRACT. Why-provenance --- explaining why a query result is obtained --- is an essential asset for reaching the goal of Explainable AI. For instance, recursive (Datalog) queries may show unexpected derivations due to the complex entanglement of database atoms inside recursive rule applications. Provenance, and Why-provenance in particular, helps debugging rule sets to eventually obtain the desired set of rules. There are three kinds of approaches to computing Why-provenance for Datalog in the literature: (1) the complete ones, (2) the approximate ones, and (3) the theoretical ones. What all these approaches have in common is that they compute provenance for all IDB atoms, while only a few atoms are requested to be explained. We contribute an on-demand approach: After deriving all entailed facts of a Datalog program, we can query for the provenance of particular IDB atoms and the structures involved in deriving provenance are computed only then. Our framework is based on terminating existential rules, recording the different rule applications. We present two implementations of the framework, one based on the semi-ring solver FPsolve, the other one based on the recent sets extension of Datalog, known as Datalog(S). We perform experiments on benchmark rule sets using both implementations and discuss the feasibility and potential of provenance on-demand.

ABSTRACT. In ontology-mediated query answering, access to incomplete data sources is mediated by a conceptual layer constituted by an ontology, which can be formulated in a description logic (DL) or using existential rules. In the literature, there exists a multitude of complex techniques for incorporating ontological knowledge into queries. However, few of these approaches were designed for explainability of the query answers. We bridge these two qualities by adapting an existing proof framework toward conjunctive query answering, based on the notion of universal models. We investigate the data and combined complexity of determining the existence of a proof below a given quality threshold, which can be measured in different ways. By distinguishing various parameters such as the shape of the query, we obtain an overview of the complexity of this problem for several DLs.

ABSTRACT. DatalogMTL is an extension of Datalog with metric temporal operators that has found applications in temporal ontology-based data access and query answering, as well as in stream reasoning. Practical algorithms for DatalogMTL are reliant on materialisation-based reasoning, where temporal facts are derived in a forward chaining manner in successive rounds of rule applications. Current materialisation-based procedures are, however, based on a naive evaluation strategy, where the main source of inefficiency stems from redundant computations.

In this paper, we propose a materialisation-based procedure which, analogously to the classical seminaive algorithm in Datalog, aims at minimising redundant computation by ensuring that each temporal rule instance is considered at most once during the execution of the algorithm. Our experiments show that our optimised seminaive strategy for DatalogMTL is able to significantly reduce materialisation times.

ABSTRACT. Query reformulation under constraints is an essential part of modern query optimizers. This paper introduces an enhancement to an interpolation-based rule-driven query optimizer that extends the space of valid rewritings for a user query in order to find better execution plans otherwise not found. The enhancement is inspired by the so-called Magic Set transformation (MST). However, in contrast with the traditional use of MST, our approach uses MST-like transformation to derive additional formulae constituting a desirable extension to the space of query plans.

ABSTRACT. The fight between the symbolic and sub-symbolic schools of AI appears to be mostly over, as there seems to be a growing consensus that AI needs the two approaches to join forces, lest we will face a new AI winter. In this talk, I will look at the fast growing field of Neuro-Symbolic AI from the point of view of rules: how neural networks are used to represent, learn and execute rules, what are (some of) the associated problems and challenges, what are the benefits from these approaches and why they are important for the future of AI. I will present solutions that have been proposed for different kinds of rules, and I will focus particularly on the case of decision rules.