Year: 2021
Type: dissertation
Abstract: Je déclare avoir respecté, dans la conception et la rédaction de ce mémoire d'HDR, les valeurs et principes d'intégrité scientifique destinés à garantir le caractère honnête et scientifiquement rigoureux de tout travail de recherche, visés à l'article L.211-2 du Code de la recherche et énoncés par la Charte nationale de déontologie des métiers de la recherche et la Charte d'intégrité scientifique de l'Université de Montpellier.Je m'engage à les promouvoir dans le cadre de mes activités futures d'encadrement de recherche. RESEARCH ACTIVITY1.1 OVERVIEW The intelligent exploitation of data poses the significant challenge of devising automated reasoning techniques able to improve access to information.Today, with data produced at high rates and in a large number of forms, exploiting available information becomes increasingly difficult.Nevertheless, in this dynamic and complex situation, there is a point of convergence: domain-specific knowledge.Data is generated by software applications, which are always proper to a specific context and domain.Hence, regardless of the many disparities, a common ground for the data produced in an application domain always exists.Formalized domain-specific knowledge gives a conceptual point of view of the domain of data, by defining the types of the objects involved, as well as the relations among them.This provides a lens through which diverse data can be looked at.It allows one to equalize databases presenting structural differences as well as to push further their value in terms of informative content.Importantly, formal knowledge can be taken into account automatically by leveraging on reasoning, in a way which is transparent for users.Modeling and leveraging knowledge through formal languages is at the essence of the Knowledge Representation and Reasoning (KR&R) field of Artificial Intelligence.However, this also concerns the fields of Databases and the Semantic Web.In the last decade, these three domains have been actively interacting on the different facets and issues related to the intelligent exploitation of data.The questions range from the foundations of reasoning, to performance issues, to the development and deployment of Web standards and solutions.Domain-specific knowledge is typically expressed by ontologies or, more generally, by rule-based languages with a logical semantics.The purpose of rules is perhaps described at best by Gottlob: "rules encapsulate transferable knowledge, arising from human experience, common sense, definitions, norms, laws, and regulations" but this also includes "knowledge that can be learned by machines" [68].Rules make one able to deal with intensionality and incompleteness in the data.This is what makes them able to enrich databases with novel and pertinent information thereby increasing their informative content and exploitability.The research activity presented here revolves around rule-based languages for reasoning on data, with a particular emphasis on existential rules.This is an expressive formalism that encompasses well known formalisms like Datalog and Horn Description Logics.The main problem we are concerned with is that of answering queries under sets of existential rules.This is commonly referred as ontology mediated query answering (or theory-mediated query answering, if we see rules as logical theories).The three main topics that will be covered are the analysis, design, and applications of rules languages for reasoning on data.Our attention has been on understanding the power of logical reasoning, in particular in the light of reusing existing database technology such as NoSQL systems.The static analysis of rule based languages serves at deriving properties that can reveal to be key for choosing the right reasoning strategy, as well as for a better reuse of rules in a collaborative setting like that of Web ontologies.The issue we have investigated is boundedness.This property concerns the recursivity of rules, and asks whether the depth of reasoning is independent on the input data.
Source: HAL
Author: Ulliana, Federico
Language: English
Topic: Semantic Web and Ontologies
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W4392492936
