Towards contextual ontologies requirements and formalization - Archive ouverte HAL Accéder directement au contenu
Pré-Publication, Document De Travail Année : 2005

Towards contextual ontologies requirements and formalization

Résumé

Research topic Domain specific ontologies are intended to provide a commonly shared understanding among several user communities. Concepts and their relationships are used to explicitly model the real world phenomena that are of interest to users of specific domain. In reality, a set of applications of a specific domain use different representations of the same real world entity due to various abstraction mechanisms, viewpoints, context, and specific interest. In the last decades, the notion of context has been applied by the AI and computer science as part of abstracting and building conceptual models for knowledge representation systems, and very recently in semantic Web and context-aware computation. In our research, the context will be adapted as meta-data to annotate ontological information elements (concepts, roles, objects, attributes, etc.) to describe their contextual information. Our specific objective is to express ontological information elements as a data set with the multiple perspectives using a modal description logics language (Modal ALCN). Problems and objectives In the dissertation, we are investigating the multi-representations problem and its impact on semantic heterogeneity for domain-specific ontology. The focal points of this work are: (i) study the requirements of the multiple representations ontologies based on the notion of context, and (ii) formalize the multi-representation ontologies by exploiting contextual description logic language that must be used to express context-dependent domain ontologies. This research was inspired by the fact that multi-representations may be an acute source for semantics interoperability and the future need for the coordination among multi-representations systems. Moreover, multi-representations systems provide users with several views of information in different contexts. Hence, constructing ontology based on context can disambiguate multi-represented objects and participate in resolving semantic heterogeneity. Moreover, by contextual ontology approach we allow local inconsistencies and enhance modularity of a set of ontologies (multiple ontologies). There are also other benefits from constructing contextual ontologies namely tracking of dynamic changes and versioning, collaborative modular development among diversified communities in the same domain. Our approach The approach is not intended for integrating a set of ontologies, but rather to preserve the semantics of those ontologies according to their local views or contexts for the purpose of accessing, filtering and querying heterogeneous sources. Our goal in this research will be to tackle the problem of multi-representation from the semantics viewpoint by providing a well-defined rich model to multi-representation ontologies as well as supporting ontology mappings (context based) when integration is not possible or infeasible. There are approaches similar to this approach namely contextualizing Ontologies or C-OWL, Distributed Descriptions Logics, and multi-context systems. In the C-OWL approach, an extension of the syntax and semantics of OWL is proposed to allow for the representation of contextual ontologies. In Distributed Description Logics, an extension of the logical formalism of Description Logics is intended for loosely coupled federations of information systems (IS), where the local IS preserves a degree of monotony. Our approach is somewhat similar to the C-OWL approach but our objectives are different. First, our approach will be dedicated to the multi-representations problem of ontological elements. Secondly, we propose constructs of modal description logics to describe the structural relationships between contexts. Third, the proposed contextual language is exploiting some modalities of modal logics to support the accessibility of information based on contexts. In brief, our approach will be dedicated to (i) study and analyze contextual ontology requirements, (ii) Extend description logics existing languages to support the notion of contextual ontologies Proposed framework The Figure below illustrates the association of information systems to their local ontologies. Each information system is supported by local ontology to gain semantically sound representation. The notion of context associated with local ontology can play the role of partitioning the ontological information. Furthermore, developing several ontologies in the same domain of interest will require ontology mapping, merging and/or alignment based on context. Multiple ontolgies can be: (i) disjoint such that no concepts are common, (ii) overlapped where concepts are commonly shared, and (ii) subsumed such that ontology is contained in another one. In this work, we consider the last two cases where ontologies overlap or subsume one another. In overlapping ontologies, ontological concept, represented by a class, may conceptually occur in other ontology as well. In all ontologies, where the concept occurs, a different context for that concept is considered. Thus, we have the problem of expressing the relationships between the different concepts in different ontologies that actually represent the same concept in different contexts (multi-representation problem). The approach works as it follows: 1. Ontologies are kept as the domain experts built them. Several ontologies are built according to the need and interests of a group of users of a specific domain. 2. Contexts are used as labelling mechanism to distinguish ontological information. 3. Formalizing concepts by exploiting proposed contextual description logics language. 4. Mappings, based on context, are defined between relevant portions of two local ontologies to express the occurrences of the same concept in different contexts. Interpretation of a concept is done according to context (s) where it occurs. The framework of the proposed approach The advantages of the proposed approach are: • No modification to the domain ontologies, since each group is responsible for its own local ontology. • The approach promotes modularity in the ontologies developed for a specific domain of interest. • Domain experts are relieved from the difficult task of reaching consensus (ontological commitments) among a set of users in the domain. • Semantically sound representation for multi-represented ontological elements, which may be valid in one context and not valid in another. Hence, inconsistent local ontologies are permitted. Contribution Our contributions to contextual ontology will be summarized in the following. • We explore and emphasize the significance of the notion of context for efficient abstraction and information space partitioning of ontological information elements. • We specify the requirements of contextual ontologies as well as their impact on future conceptual system modelling and systems interoperation. • We devise formalism for contextual ontology by extending language constructs of modal description logic family (ALCNM) to cope with the problem of the multi-representation ontological concepts.
Les ontologies de domaine sont prévues pour fournir une compréhension généralement partagée parmi plusieurs communautés d'utilisateur. Les ontologies comportent des concepts, des instances, des liens entre concepts, et des axiomes. Ces derniers sont employés pour modéliser explicitement les phénomènes du monde réel d'un domaine spécifique. En réalité, un ensemble d'applications d'un domaine spécifique emploie différentes ontologies. Par conséquent, les différentes représentations de la même entité du monde réel existent en raison de divers mécanismes abstraits, des points de vue, du niveau de détails, et d'intérêt de l'utilisateur. Dans les dernières décennies, la notion du contexte a été utilisée par l'Intelligence Artificielle (IA) et l'informatique en tant qu'élément pour abstraire et établir les modèles conceptuels, systèmes de représentation de la connaissance, et très récemment dans les applications contextuelles. Dans notre recherche, l'information contextuelle est prise en compte en adoptant la notion du contexte et en l'intégrant dans la description des éléments ontologiques (concepts, rôles, objets, attributs, etc. ). Notre objectif est d'exprimer les éléments ontologiques dépendants du contexte comme un ensemble de données avec des perspectives multiples en utilisant les logiques de description. Plus précisément, nous étudions une approche pour décrire les ontologies qui partagent l'entité du monde réel avec des représentations multiples. Notre recherche est focalisée sur deux secteurs : (i) les besoins des ontologies avec des représentations multiples et (ii) un formalisme pour exprimer les ontologies de domaine qui peuvent exister dans différents contextes. Cette recherche a été motivée par le fait que les représentations peuvent changer selon différents critères tels que le point de vue, la granularité d'information disponible, la classification des concepts, le temps, l'espace, etc. En fait, les repréeentations multiples des phénomènes du monde réel sont de plus en plus standardisés
Fichier non déposé

Dates et versions

hal-01469721 , version 1 (16-02-2017)

Identifiants

  • HAL Id : hal-01469721 , version 1

Citer

Ahmed Arara. Towards contextual ontologies requirements and formalization. 2005. ⟨hal-01469721⟩
80 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More