Ontonym: A Collection of Upper Ontologies for Developing Pervasive Systems - Leitura de Artigo

 ABSTRACT

Pervasive systems present the need to interpret large quantities of data from many sources.

[Integração de fontes de dados]

Context models support developers working with such data by providing a shared representation of the environment on which to base this interpretation. This paper presents a set of requirements for a context model that addresses uncertainty, provenance, sensing and temporal properties of context.

[Dimensões contextuais que estou tratando no CKG: proveniencia, temporal, localidade e identidade. Outras dimensões podem ser Incertezas e "Sensores" e isso depende do domínio do KG e/ou aplicação]

1. INTRODUCTION

Pervasive computing is an evolution of the desktop computing paradigm, whereby almost any object, from home furnishings and appliances, to cars, to clothing, even to coffee mugs and credit cards can be embedded with sensing and processing capabilities [1].

[IoT]

Through networks of these devices, information about people and their surrounding environment is combined and used to provide personalised services across application domains as diverse as assisted living, environmental monitoring, and ambient information systems

[Informações que trafegam por essa rede podem ser usadas para personalizar o serviço]

The information relevant to a particular system or application is referred to as its context. Dourish describes context as a “slippery notion” [2]; it is highly domain specific, and has given rise to definitions too restrictive to obtain consensus and too broad to be meaningful.

[Não existe uma definição consensual de contexto]

While the lack of an accepted definition precludes the specification of a “complete” model of context (that is, a symbolic model of the environment against which developers can write software), an extensible model that describes key pervasive computing concepts provides a basis for sharing data and supporting interactions between systems.

[CKG é um modelo simbólico extensível para agregar novas dimensões contextuais que permite a construção de um KG para busca exploratória ciente de contexto]

We then describe Ontonym, a set of upper ontologies for pervasive computing ... , we regard upper ontologies as describing high-level, but domain specific (i.e., pervasive computing) concepts, as opposed to top-level ontologies, which model the fundaments of the world.

[Diferença entre ontologias TOP e UPPER?]

2. MODELLING CONTEXT

Strang and Linnhoff-Popien [4] and Henricksen et al. [5] set out several requirements for a context model, covering technical requirements of the modelling technique, capturing the quality of data, and supporting the representation of past and future states.

[Olhar essas referencias]

[4] T. Strang and C. Linnhoff-Popien. A context modelingsurvey. In Proceedings of the Workshop on Advanced Context Modelling, Reasoning and Management as part of UbiComp 2004 - The Sixth International Conference on Ubiquitous Computing, Nottingham, UK, September 2004.
[5] K. Henricksen, J. Indulska, and T. McFadden. Modelling context information with orm. In Proceedings of On the Move to Meaningful Internet Systems 2005: OTM Workshops, pages 626–635, 2005.

2.1 Uncertainty

... model incomplete, ambiguous, and imprecise information. We put these under the umbrella heading of uncertainty, using the definitions of Henricksen et al. [6], as they relate to the value of an attribute belonging to an entity.

[Dimensão associada a propriedades e não relações?]

Incompleteness occurs when not all information about the attribute is known;

[No caso de KG incompletude é uma caracteristica de todo KG (OWA) e não uma dimensão. Mas as regras podem especificar o uso de LCWA para algumas propriedades ou relações em alguns contextos]

ambiguity when two or more data sources provide contradictory information about the attribute;

[No caso do KG contraditório também é uma característica em potencial e informações contraditórias podem conviver no KG desde que contextualizadas]

and imprecision when the attribute’s value is an approximation of the real world.

[Esse é o caso que vou efetivamente tratar como dimensão contextual]

2.2 Provenance

Provenance is concerned with being able to determine the origin of a piece of data by recording from where data was sourced, and the role played by intermediate components in its derivation [7]. For any value in the data model, it should be possible to trace the complete transformation history of the data from sink to source.

[A entidade fonte / origem e todo os seus atributos. A proveniência pode ser por statement (aresta), por entidade (nó) ou para um subgrafo]

2.3 Sensing

As information in pervasive computing systems is largely discovered through sensors observing the environment ... The representation of a sensor can be used in tracking provenance, and, where aspects of readings or its meta-data are constant, these properties can be modelled as part of the sensor rather than repeated across each of its readings.

[Este contexto é específico de sistemas pervasivos. Mas o equipamento/método pode ser um contexto mais geral aplicável a outros domínios. Por exemplo, um município pode ter uma imagem de satélite e nessa relação ser qualificado a entidade satélite e com suas propriedades]

2.4 Temporal Properties

There are four temporal properties that affect the interpretation of context, and may inform the design of the system that manages and provides access to it: dynamism, temporal dimension, observation time, and sampling period. ... The temporal dimension of data is concerned with whether data describes past or planned state (e.g., a prior event or a scheduled meeting), while observation time records when the data was sampled. Finally, sampling period describes the frequency of a sensor’s observations.

[Eventos temporais contínuos ou instantâneos, passado, presente, futuro]

2.5 Modelling Capabilities

distributed composition states that no central authority should be responsible for the administration of a context model and its data; partial validation requires that it is possible to validate the structure and content of knowledge against a model, even if complete knowledge is not available in a single place at the same time; and a level of formality ensures that each participant in an interaction shares the same interpretation of the data exchanged.

[1. A manutenção do modelo contextual não é centralizada. No caso do CKG, o engenheiro de conhecimento mantém]
[2. Validação parcial pq o dado é incompleto. No caso do CKG, as dimensões podem ser adicionadas gradativamente]
[3. Formalismo para garantir a interpretação de ambos os agentes. No caso do CKG, nada fica implícito, o contexto é explicito]

We therefore argue that such requirements can be supported by using a modelling technique that supports meta-modelling, i.e., the ability to annotate statements in the model.

[Reificação e Hiper-relacional]

3. THE ONTONYM ONTOLOGIES

Although context is an application specific notion, context-aware applications exhibit overlapping data requirements; the most common being the need to represent time, location, people (or identities), and events (or activities) [11].

[Eventos seriam contextos também ou afirmações que podem ter contexto temporal, localidade, ...]

The ubiquity of these concepts in the literature suggests their role in forming upper ontologies in a context model, from which application specific contexts can extend.

3.1 Time

The ISO 8601 standard for date and time representations covers lexical representations of Gregorian dates, time of day, and time intervals – a subset of which is adopted by XML Schema.

[Literais do tipo data / tempo]

.. relationships between intervals (during, starts, finishes, before, after, meets, and equals) and their inverses for a total of 13 (equals has no inverse). the W3C OWL-Time ontology provides a comprehensive vocabulary;

[As operações poderiam ser suportadas por regras]
[OWL-Time: www.w3.org/TR/owl-time]

3.2 Location

Physical positions, such as those generated by the Global Positioning System (GPS), take the form of a 2D or 3D numeric array. Symbolic positions,... describe locations using human-friendly descriptive names, such as “Coffee Area”, that may be organised into a hierarchy of granularities (e.g., coordinate, room, and building).

[E ainda teriam classes de lugares como igrejas, praias, cafeterias, cemitérios]

Ontonym’s location ontology is an implementation of Ye’s model. Spaces are represented using a combination of SymbolicRepresentation, GeometricRegion, and RelativeLocation ontology classes. The model defines four types of spatial relationship: containment, adjacency and overlap are as per their suggested meaning, while connectedness is a particular case of adjacency, where it is possible to pass from a space to its adjacent space. Coordinates are defined with reference to a coordinate reference system (CRS), and translations from one CRS to another ...

[Não usaram uma Ontologia conhecida como a GeoSPARQL]

3.3 People

Excluding application-specific data, the types of informa tion about a person used in pervasive systems can be broken down into four categories: identity, device ownership, personal details, and social relationships. Identity allows applications and services to differentiate one user from another. .... OWL’s URIs provide the notion of identity. We adopt the Friend-of-a-Friend (FOAF).

[A entidade Pessoa não é um contexto mas a sua Identidade é. A ontologia FOAF poderia ser usada para selecionar possível identificadores.]

3.4 Sensing

Ontonym’s sensing ontology is concerned with the description of sensors and the data they generate. We represent the characteristics and uncertainty of sensed data by using a quality matrix.

[Contexto de Incerteza e de Equipamento/Método]

Distances are described using the Measurement Units Ontology, and an OWL representation of the Unified Code for Units of Measure.

[The MUO ontology: idi.fundacionctic.org/muo]
[The UCUM ontology: unitsofmeasure.org]

3.5 Provenance

The provenance ontology in Ontonym models three things: the creator or author of data, the time at which data is created or modified, and the data from which new data is derived.

[Datas podem fazer mais de mais de um contexto]

3.6 Events

The event ontology provides a means of describing activities that have (at least) a temporal dimension. ... A further two classes SpatioInstantEvent and SpatioIntervalEvent represent temporal events with an associated location. The event ontology also defines the Role class and containsRole property to identify roles that are played by entities ....

3.7 Minor Ontologies

Ontonym contains a couple of additional minor ontologies.

4. PRINCIPLES OF ONTOLOGY EVALUATION

Representing knowledge is a highly subjective process – what is fit for one purpose may not be for another.

4.1 Design Principals

Extensibility requires that new classes and properties should be easily integrated with an ontology to meet the requirements of a given application domain, with minimal modification.

[Novas dimensões podem ser adicionadas]

4.2 Content

Clarity relates to the specification of concepts in an ontology and requires that for each concept, all necessary and sufficient conditions that distinguish it from another concept are represented. Consistency requires that terms map to their real-world understood meanings, while syntactic correctness ensures that no incorrect terms or documentation appear. ... Finally, conciseness is the don’t-repeat-yourself principle of ontology engineering; existing terms should be reused where possible and redundancy should be eliminated.

[Critérios de qualidade do conteúdo]

4.3 Evaluation for Purpose

Determining from a set of ontologies which is best suited for a particular purpose is difficult.

... As a halfway measure, we propose that ontologies for sensing may be evaluated by applying them to examples from live sensor feeds or published data sets.

5. AN INITIAL EVALUATION OF ONTONYM

5.1 Design Principals and Content

5.2 Modelling Requirements

Use of an RDF-based model also supports meta-modelling through reification. This allows a single statement in the model to be associated with any number of meta-data statements (e.g., representing ownership, security, or assertion of error).

5.3 Comparison with Related Ontologies

The Context Ontology Language (CoOL) [29] focuses on providing a comprehensive description of services and context interaction, but does not fully represent people or time,

[29] T. Strang, C. Linnhoff-Popien, and K. Frank. CoOL: A Context Ontology Language to enable Contextual Interoperability. In Proceedings of 4th IFIP WG 6.1 International Conference on Distributed Applications and Interoperable Systems (DAIS2003), LNCS 2893, pages 236–247, Paris, France, November 2003

6. CONCLUSION