The STAR:dust conceptual model: Semantic Travel Across Resources (STAR) with the aim of designing unified support tools (dust) to help travelers in their navigation
Contact e-mail: irene.celino # cefriel.it (main contact)
emanuele.dellavalle # cefriel.it
francesco.corcoglioniti # cefriel.it
Application
General purpose and services to the end user
STAR:dust is a conceptual model aimed at designing and specifying the navigation, i.e. the "travel" that web users undertake while surfing through resources. It provides a thorough conceptualization that can be used as "application ontology" (in a Model-driven Architecture approach) for software tools (called "vehicles" in the metaphor of travel) that support the navigation and the presentation of resources.
Functionality examples
The STAR:dust model is used to design:
- the generation of pages: what information about a resource should be displayed and how it should be presented to the users;
the navigational paths: how a user can navigate through resources by exploiting the links between resources; STAR:dust can be used for designing pre-defined paths to be followed across information;
- multiple views on information: the model allows the definition of different "travels" across resources, e.g. taking into account users' profiles or preferences.
For example, in the presentation model, we describe the different options to visualize a long text: the text could be displayed in full at the center of the page; or it could be abbreviated (the first few words followed by dots) leaving the rest of the text "behind" a hyperlink or a button; finally, it could be inserted in a box or frame with scrollbars, in order to present all the details without taking up too much space.
Application architecture
The STAR:dust conceptual model specifies the "navigation and presentation semantics". The resulting vocabulary/ontology, however, is not useful per se, but it is used to strongly decouple the editing of contents from their visualization.
For example, we assume that the contents about a specific domain (e.g., artists and artwork of a museum) are edited by domain experts and provided/translated into a machine-readable format. Once we have this domain knowledge base, we can design their visualization by mapping between the domain ontology and the STAR:dust Travel model.
Finally, a tool like SOIP-F (http://seip.cefriel.it), taking as input both the domain knowledge and the mappings, makes lever on the STAR:dust model and produce a way to present and navigate across contents.
Special strategies involved in the processing of user actions
The links and pages generation are based on the mappings between the travel model and the domain-specific ontology; moreover, this can be enhanced by exploiting the semantic descriptions of the users' profiles and preferences.
Integration between vocabulary-linked functions and other application functions
We designed the STAR:dust model with the aim of enabling the implementation of model-driven applications that takes STAR:dust as "application ontology" (see SOIP-F at http://seip.cefriel.it); this approach makes the application capable to generate the pages starting from the model.
Additional references
SOIP-F (Semantic Organizational Information Portal framework) is a framework for building portals enhanced by semantics. Several domain specific portals were built on top of SOIP-F and are available on-line and listed on the web at http://seip.cefriel.it. Some publications about STAR:dust and SOIP-F are available on the web at http://swa.cefriel.it/Publications#soip-f_publications.
Vocabulary
Title
The top-level vocabulary is the STAR:dust conceptual model. The Travel model is the main vocabulary that is made up of three parts: the navigation model, the access model and the presentation model.
General characteristics (size, coverage) of the vocabulary
The STAR:dust model is made up of the main seven primitives (Vehicle, Traveler, TravelType, HyperEnvironment, TravelModel, TravelObject and Mapping) and their relations. The three parts of the Travel model are three ontologies, partially defined ad hoc (e.g., most of the access model) and partially referring to shared and wide-spread models like SKOS and Dublin Core vocabulary: the navigation model, the access model and the presentation model.
However, each application based on STAR:dust will define and exploit mappings between the Travel model and the domain-specific ontologies. In our running applications, each portals has its own ontology and we experimented both limited and very huge ontologies (with millions of triples). Domain ontologies make use of hyperonymy/hyponymy, meronymy/holonymy (part-of relation), multiple wordings (homonymy/pseudonymy/synonymy) and generic semantic relationship whenever needed.
Language(s) in which the vocabulary is provided
STAR:dust and the Travel model are not multilingual, since they are used by the software (and they don't have to be visualized to the users). The domain-specific ontologies however can have labels in multiple languages to allow the display of information in the language of the user.
Vocabulary extract
Navigation model:
- skos:related is used to define the connection between the current resource and other resources that are somehow similar or on the same subject;
- skos:broader/skos:narrower are used to represent the connections between the current resource and those resources that are at a higher/lower level of complexity;
- skos:relatedPartOf (part-of relation) represents the containment connection between the current resource and its parts (e.g., the relation between a section and its sub-sections);
- skos:Concept is used to represent the "element", i.e. every "place" where it is possible to go and the portion of information that is relevant for the navigation.
Access model:
- axs:Home is the landmark indication, i.e. the denotation of specific resources that can be taken as reference for navigation;
- axs:prev/axs:next relations are the connections between the current resource and those resources that are immediately before/after in a specific path;
- axs:up/axs:down relations are the connections between the current resource and those resources that are immediately above/below in a specific ordered list or hierarchy.
Presentation model:
It contains a lot of classes and properties to model all the characteristics of knowledge visualization. The resulting ontology is composed of both existing primitives coming from popular and shared models (e.g., properties like dc:title, dcterms:image or skos:symbol, skos:prefLabel and skos:altLabel) and other building blocks we modeled explicitly to represent e.g. the features useful for visualization functions mentioned in section 1 (pres:hasText and its sub-properties pres:hasFullText, pres:hasShortText and pres:hasSlidebarText)
Machine-readable representation of the vocabulary
The Travel model is modeled in RDF/OWL and all the domain ontologies used in running applications were also expressed in RDF/OWL.
Some sample triples from the access ontology of the Travel model:
<owl:ObjectProperty rdf:ID="next"> <rdfs:label>next</rdfs:label> <rdfs:comment>link to the subsequent resource</rdfs:comment> <rdf:type rdf:resource="&owl;TransitiveProperty"/> </owl:ObjectProperty> <owl:ObjectProperty rdf:ID="prev"> <owl:inverseOf rdf:resource="#next"/> <rdfs:label>previous</rdfs:label> <rdfs:comment>link to the previous resource</rdfs:comment> </owl:ObjectProperty>
Software applications used to create and/or maintain the vocabulary, features lacking for the case
The vocabulary maintenance is performed through an RDF/SKOS/OWL editor.
Structure of the database used to currently manage the vocabulary
We use Sesame repositories with a MySQL backend to store the knowledge bases in RDF format using Sesame pre-defined structures (therefore we didn't need to define any table structure).
Standards and guidelines considered during the design and construction of the vocabulary
Generally, we use Methontology (http://webode.dia.fi.upm.es/Asun/SSS97.ps) to build OWL ontologies. In the modeling of our SKOS-based ontologies, we made also use of the "Quick Guide to Publishing a Thesaurus on the Semantic Web" (http://www.w3.org/TR/swbp-thesaurus-pubguide/) and the "SKOS Core Guide" (http://www.w3.org/TR/swbp-skos-core-guide/).
Management of changes
The vocabulary maintenance is performed manually.
Additional references
Some publications about STAR:dust and SOIP-F are available on the web at http://swa.cefriel.it/Publications#soip-f_publications.
Some SOIP-F implementations are (cf. http://swa.cefriel.it/SOIP-F):
- Semantic-based Healthcare Information Portal
It is a Semantic Navigation engine built upon the STAR:dust model with the aim of providing General Practitioners with a system to search and navigate medical literature by leveraging on the semantics of the medical articles, expressed by a medical ontology. We have two different implementations, one developed for the COCOON project (FP6-507126, http://seip.cefriel.it/sir-demo), which uses a conceptual indexing engine and a medical ontology provided by the TeSSI suite produced by L&C, and another one (http://seip.cefriel.it/ship) that uses PubMed bibliographic references as content source and MeSH taxonomy as domain ontology.
- Virtual Museum of Contemporary Art portal It aggregates data (in Italian) about artworks and artists from different real museums, letting virtual visitors travel across resources with different "vehicles": a thematic trail vehicle, that offers a generic view on information, letting users to have a guided journey across artworks of a particular artistic movement or period, and a detailed investigation vehicle, that offers a specific view on information, by giving details about single resources (e.g., the work of a particular artist).
- Semantic Web Virtual Lesson It's a portal built as a unifying view on different material taken from some presentations about fundamentals, technologies and applications of Semantic Web; it allows end-users to move from a presentation to another and from a lecturer to another just following links to semantically related slides, as if it was a single lesson.
Vocabulary mappings
Mapped vocabularies
We use a mapping approach to put in relation the Travel model with the domain-specific ontology to build the single domain specific application. For example, in the virtual museum portal, we map between the navigation/access/presentation models and the ontology of art and artists.
Extracts of Mappings
We give hereafter some simple examples for the Virtual Museum portal.
- Mapping between domain ontology and navigation model
if an Artist painted an Artwork --> Artist skos:relatedHasPart Artwork
if a Chapter is about an Artist and describes an Artwork --> Chapter skos:related Artist, Chapter skos:related Artwork
- Mapping between domain ontology and access model
if a ThematicTrail contains a Chapter -->
ThematicTrail axs:down Chapter, Chapter axs:up ThematicTrail
if Chapter-1 is before Chapter-2 --> Chapter-1 axs:next Chapter-2, Chapter-2 axs:prev Chapter-1
- Mapping between domain ontology and presentation model
if an Artwork is represented by an Image --> Artwork skos:symbol Image
if an Artist is described by his Biography --> Artist pres:hasFullText Biography
Types of mapping used
We use different kinds of mapping between the Travel model and the domain ontology. Generally, mappings actually match any kind of (sub)graph made with the domain ontology with any kind of graph made with components of the 3 STAR:dust models. For the simplest cases, SPARQL CONSTRUCT queries are used to perform those mappings.
Additional references
See http://seip.cefriel.it for more information. Details about SOIP-F can be also found in some publications about its implementation, available on-line at http://swa.cefriel.it/Publications#soip-f_publications.