WWW.BOOK.DISLIB.INFO
FREE ELECTRONIC LIBRARY - Books, dissertations, abstract
 
<< HOME
CONTACTS



Pages:     | 1 |   ...   | 11 | 12 || 14 | 15 |   ...   | 16 |

«OBJECT DATABASES AND THE SEMANTIC WEB A THESIS SUBMITTED IN FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY. ING. JAKUB ...»

-- [ Page 13 ] --

{ om:hasLocation, om:hasThumb } rdfs:range xs:anyURI.

{ om:hasNum, om:hasImgNum } rdfs:range xsd:Integer.

om:hasAddress rdfs:range om:Address.

om:created rdfs:range om:Products.

The typing properties used here belong in the RDF/S vocabulary — their stricter SODA counterparts are only used for

–  –  –

6.4.5 EXAMPLE OF OM DATABASE CONTENTS Figure 6.5 shows three objects in the OM database that correspond to the given schema — one person, an article and an image. The emphasized nodes of the graph show SODA concepts — two class instances, one collection and one tuple. They are all SODA typed and belong in the soda:Thing class, therefore they must have all properties prescribed by their type definition. Moreover, the values of these properties must be correctly typed. One important detail is that the collection of products is polymorphic and stores both article and image objects.

Semantic Web as an Object-oriented Database 60 Another feature to notice is the new omi: namespace that serves for storing machine-generated urirefs for OIDs. These are automatically assigned to any class instances created by the user. Collection and tuple concepts are represented by blank nodes without unique identifiers.

Notice that from the RDF perspective, the pair of inverse relationships created and createdBy is redundant, since RDF triples can be read in both directions (indexed by subject and by object). On the other hand, in an object-oriented system they improve the performance and allow better traversability of the object graph.

6.4.6 EXTENDING THE EXAMPLE Now the model can be made to cooperate with other RDF vocabularies in a very simple yet powerful

way through several subclassing declarations that connect the om concepts in global class hierarchies:

xmp:Nickname rdfs:subClassOf om:hasTitle.

xmp:Thumbnail rdfs:subClassOf om:hasThumb.

om:createdBy rdfs:subClassOf dc:creator.

om:hasTitle rdfs:subClassOf { dc:title, rss:title }.

om:hasAbstract rdfs:subClassOf { dc:description, rss:description }.

om:Product rdf:type rss:item.

om:hasLocation rdfs:subClassOf rss:link.

where xmlns:rss="http://purl.org/rss/1.0/” xmlns:xmp="http://ns.adobe.com/xap/1.0/” xmlns:dc=”http://purl.org/dc/elements/1.1/” If the maintainer of our example database decided to install a simple RDF reasoner and a RDF server,

the above statements would result in opening the online magazine database to the following systems:

XMP – Extensible Metadata Protocol: The hasTitle and hasThumb properties can now be imported from PDF or Photoshop files upon insertion in the database since Adobe XMP metadata is embedded in those files and the values of two different properties are identified by the OM system.

RSS – Remote Site Syndication: RSS information is now automatically generated by every Product object in the database because it also becomes a RSS Item object. If the RDF server took the result, other sites could display brief descriptions and titles of the images and articles and other users could download the RSS feeds into their desktop readers to check for updates in their favorite magazine.

DC – Dublin Core Metadata: By publishing Dublin Core information about the creator, title and description of products in the magazine, the whole database opens up to third party semantic searches and indexing, because it suddenly uses a widely accepted metadata protocol.

Semantic Web as an Object-oriented Database 61

–  –  –

7.1 MINING OBJECTS FROM RDF/S In this context, mining objects means finding object-oriented structures in arbitrary RDF/S graphs, extracting objects from RDF graphs and storing them in an object database. This can be accomplished by either accepting the existing RDF/S schema, or building a new schema based on regularities encountered in the graph, or taking an existing OODB schema. This section focuses on the first approach and contains guidelines that are also useful in the other two approaches.

7.1.1 MOTIVATIONWhat are some reasons for mining objects from RDF data?

Structured processing — object-oriented databases like Zope (http://www.zope.org/) contain application and Web servers so once the data is stored in such systems, it can be very easily extended with methods, published and manipulated. Relational databases can also structure data, but cannot take full advantage of RDF Schema information.

Object-oriented storage — RDF graphs could be directly stored in OODB systems. Performance problems with storing objects in tables have been addressed by several approaches, like using sparse arrays (Caché object database).

Type checking — once the object-oriented schema is defined, it is possible to automatically validate the required structure of arbitrary RDF data. Algorithms that process the result do not need to do so much type checking — this is similar to having a DTD for XML files.

Deductive extensions — looking at the advantages from a database viewpoint, an object database that can access RDF data can use RDF entailments and therefore gains certain deductive features.

7.1.2 OVERVIEW OF EXTRACTED STRUCTURES Most objects can be extracted automatically by imposing the SODA-interpretation on RDF data, converting RDF Schema statements into corresponding SODA statements and employing machine entailment but several minor adaptations of the RDF graph can extend this process and extract even more objects, although the resulting schema can be quite unstructured.





Semantic Web as an Object-oriented Database 62 To extract object data from RDF, a list of basic object-oriented concepts and their RDF counterparts is needed. These concepts, corresponding to the ODMG standard [CB00] and G2 Concept Definition Language [HM00], are listed below.

Objects — All data in an OODB are stored in objects. An object has a unique OID (uriref in RDF) and it contains a tuple or a collection of attributes or references.

All RDF nodes with a uriref are considered objects.

Datatypes — Most OODB models build their complex types on top of a set of elementary datatypes. These cannot be further decomposed and their semantics is fixed. This approach is equivalent to RDF T-interpretations.

In both RDF and an object database, an instance of a datatype is bound to one type and its value is physically contained in the data node. Datatype nodes are converted to atomic values.

Attributes and relationships — A collection or a tuple can either contain or reference a value.

This is important for modeling, data storage and update semantics.

When the value of a RDF attribute is a uriref node, it is referenced, and when its value is either a datatype or a blank node with a single reference to it, its value is embedded within the parent object. Blank nodes with multiple references are handled in a way similar to JDO Second Class Objects [Craig03] — an identical copy of all edges starting in this node is stored for every incoming edge.

Tuples — A tuple is an elementary modeling concept; a structure with attributes labeled by property urirefs.

Every blank RDF node that has at least one non-collection attribute can be considered a tuple.

Adding a uriref to the node makes it a tuple object.

In cases where the RDF graph contains several triples that assign different values to the same attribute of a node, an appropriate collection of the most general type is substituted for these attributes in the object-oriented schema.

Collections — RDF has collection vocabulary without adequate formal semantic restrictions. In OODBs, collections are often typed and they are used to model extents or 1-to-N relationships.

Any blank RDF node with collection attributes (eg. rdf:_1) can be considered a collection. A RDF collection with a uriref is reified to become a class instance with an embedded collection. RDF can have objects that act as both tuples and collections, similarly to G2 CDL [HM00], and the formal SODA model respects this.

Types — In most OODBs, an object has exactly one type that specifies its internal structure.

However, a RDF node can have no types or multiple types. Having no type is equivalent to having the most general type (soda:Thing) and having multiple types is similar to the concept of multiple inheritance and object roles. Nodes that conform to the structure prescribed by their types (through rdf:type, domain and range properties of object attributes) are labeled as "strongly typed" (soda:type) in the model and their class becomes a SODA concept.

Inheritance — In both RDF and OODBs, inheritance is a fundamental tool for type hierarchies.

The RDF definition of inheritance as a subset relationship on elements of the universe is suitable for its database counterpart, because it allows multiple inheritance and preserves the notion of "strongly typed" objects. SODA inheritance is applied to SODA concepts connected by the rdf:subClassOf attribute.

With these guidelines an object-oriented structure can be extracted from any RDF graph. Depending on the schema and organization of the graph, the format of the resulting data can be either quite loose Semantic Web as an Object-oriented Database 63 with many extra attributes or strongly typed — the SODA model allows both. Many custom RDF vocabularies are quite simple and place strong restrictions on their data, so most practical Semantic Web applications provide strictly structured RDF graphs suitable for OODB processing.

7.2 ACCESSING GRAPH DATA

7.2.1 ACCESSING THE RDF GRAPH For the purposes of reasoning, a RDF graph is often presented as a set of facts (triples — binary predicates). When retrieving binary predicates for deductions, the most common request is to find a set of triples with a constant predicate — in Prolog syntax, this would be something like parent(X,oid1) or supervises(X,Y). This means that from the viewpoint of retrieving data, a RDF graph can be accessed randomly — the access point is the label of an edge, therefore the graph does not need to be connected and it is not important whether all its nodes are reachable. For representing knowledge bases in a database, this approach leads grouping the data by predicates.

While this is common from the deductive point of view, it presents an obstacle for viewing the RDF graph as an object database that usually has a different structure.

7.2.2 ACCESSING AN OBJECT DATABASE Object database (without a query language) are accessed differently. In an OODB, information is physically grouped by objects. An object is accessed as a whole, and in contrast to the deductive approach, it is unusual to retrieve all occurrences of a given attribute. In Prolog syntax, accessing the whole object can be expressed as X(oid1,Y), which can be translated as "find all attributes X of object obj1 and read their values Y". The access point into the database graph is the label of a node. In a typical object database, only a limited number of nodes can be accessed directly and other nodes need to be reached by traversing the edges (references/attributes). Most common entry points are extents — for a given type, an extent is an object that stores a collection of its instances.

The practical result is that an object database is accessed from several nodes by traversing the edges of the object graph. In contrast to the Semantic Web approach, the direction of an edge is important and the whole graph must be reachable from its access points.

7.2.3 A COMMON ACCESS MODEL To access the RDF graph as an object database, the whole graph needs to be reachable from certain access points. Figure 7.1 shows part of a RDF graph ("Nathaniel majors in philosophy"). To make such graph fully reachable, four changes to the original structure need to be made (three of the

changes are emphasized and my: namespace contains user defined objects and attributes):

Semantic Web as an Object-oriented Database 64

–  –  –

Adding type extents. In a typical OODB, every object needs to be part of at least one extent.

For types in the database, extents are collections that are directly reachable from the system catalog (the only initial access point into the database), which in effect makes all the objects in the database accessible. In RDF setting, extent nodes could actually be identified with type nodes (oodb:PersonExtent and my:Person).



Pages:     | 1 |   ...   | 11 | 12 || 14 | 15 |   ...   | 16 |


Similar works:

«Charakterisierung der Rolle zellulärer Proteine bei der Transkription und Replikation von Papillomviren Inaugural–Dissertation zur Erlangung des Doktorgrades der Mathematisch-Naturwissenschaftlichen Fakultät der Universität zu Köln vorgelegt von Manuela Rehtanz aus Leverkusen Berichterstatter: Prof. Dr. Herbert Pfister Prof. Dr. Karin Schnetz Vorsitzender: Prof. Dr. Dietmar Schomburg Beisitzer: Dr. Rolf Kaiser Tag der Disputation: 14.07.2004 Den Menschen, die mich begleiten. Das Erkennen...»

«Landnutzung, Biodiversität und Naturschutz Eine Fallstudie über die Nutzung privater Waldfragmente der Mata Atlântica im Biosphärenreservat Grüngürtel der Stadt São Paulo, Brasilien Inauguraldissertation zur Erlangung der Doktorwürde der Philosophischen Fakultät der Universität zu Köln vorgelegt von Jutta Schmidt-Silva Machado Institut für Ethnologie Universität zu Köln Köln, im April 2012 Die vorliegende Arbeit wurde von der Philosophischen Fakultät der Universität zu Köln im...»

«Mythos – Geschlecht – Medien Die Nibelungen Ein kulturhistorischer Vergleich Dissertation zur Erlangung des akademischen Grades doctor philosophiae (Dr. phil.) eingereicht an der Philosophischen Fakultät III der Humboldt-Universität zu Berlin von Simone Schofer geb. am 3.2.1975 in Worms am Rhein Präsident der Humboldt-Universität zu Berlin Dekan der Philosophischen Fakultät III Gutachterinnen: 1. Prof. Dr. Christina von Braun 2. Prof. Dr. Inge Stephan Datum der Promotion: 25. 04. 2007...»

«DIPLOMARBEIT Titel der Diplomarbeit „Sag an mein Freund, die Ordnung der Unterwelt.“ Das Gilgamesch-Epos in Hans Henny Jahnns „Fluß ohne Ufer“ Verfasserin Gianna Zocco angestrebter akademischer Grad Magistra der Philosophie (Mag. phil.) Wien, 2009 Studienkennzahl lt. Studienblatt: A 393 Studienrichtung lt. Studienblatt: Vergleichende Literaturwissenschaft Betreuer: Ao. Prof. Dr. Norbert Bachleitner Inhaltsverzeichnis 1. Annäherung an ein „Riesenepos“ 2. Das Gilgamesch-Epos 2.1 Was...»

«Limitations Upon Legitimate Authority to Suspend and Expel K-12 Public School Students: A Moral Analysis Robert F. Ladenson Illinois Institute of Technology Chicago, IL 60616 The topic of morally legitimate disciplinary authority in K-12 public schools, despite its immense importance, has received little, if any, focused attention from moral philosophers and philosophers of education. This is understandable given nearly universal agreement that disciplinary authority of staff in any school, at...»

«Früherkennung von Entwicklungsstörungen durch Erzieherinnen: Überprüfung der Gütekriterien des Dortmunder Entwicklungsscreening für den Kindergarten (DESK 3-6) vorgelegt von Judith Flender als Dissertation zur Erlangung des Grades einer Doktorin der Philosophie (Dr. phil.) in der Fakultät Rehabilitationswissenschaften der Universität Dortmund Dortmund Betreuer: Prof. Dr. Heinrich Tröster Betreuer: HD Dr. Gerd Mannhaupt Meinen Eltern Danksagung Ganz herzlich möchte ich mich bei allen...»

«B er.K _r.N U r n b g.E n t. G a.la.th e a. 6/4 N ü r n b e r g 1990 Deutsche Naturforscher und Entomologen in Südamerika III. Dr. Fritz Müller Stefan Kager Zusammenfassung: Der Name Fritz Müller steht gleichermaßen für eine enorm gründliche und vielseitige wissenschaftliche Erfor­ schung der Natur Brasiliens, für eine philosophische Betrachtung und Untermauerung von Wissenschaftstheorien, wie auch für die Schwierigkeiten, die viele deutsche Einwanderer in Südamerika im letzten...»

«DIPLOMARBEIT Titel der Diplomarbeit „Erinnerung, Gedächtnis und Identität in ausgewählten Werken Laura Goodman Salversons und Kristjana Gunnars‘“ Verfasserin Elisabeth Paleczek angestrebter akademischer Grad Magistra der Philosophie (Mag.phil.) Wien, 2013 Studienkennzahl lt. Studienblatt: A 393 Studienrichtung lt. Studienblatt: Diplomstudium Vergleichende Literaturwissenschaft Betreuerin / Betreuer: Dr. Barbara Agnese Inhaltsverzeichnis Danksagung 1 1. Ziel dieser Arbeit 2 2....»

«DIPLOMARBEIT Titel der Diplomarbeit Dankbarkeit und Körper Dankbarkeit für den Körper Verfasserin Jovita Susanna Tatschl angestrebter akademischer Grad Magister der Philosophie (Mag.phil.) Wien, im August 2012 Studienkennzahl lt. Studienblatt: A 307 Studienrichtung lt. Studienblatt: Kulturund Sozialanthropologie Betreuer: Mag. Dr. Hermann Mückler ao. Univ.-Prof. Abstract Der Körper ist ein stiller Begleiter des Menschen, der oft, solange er funktioniert, wenig Dankbarkeit und...»

«Gender Transitivity in Three Dramatic Works by Robert Schumann: the Szenen aus Goethes Faust (WoO 3), Genoveva Op. 81, Das Paradies und die Peri Op. 50 By Claudine Jacques Department of Music McGill University, Montreal February 2011 A thesis submitted to the Faculty of Graduate Studies and Research in partial fulfillment of the requirements of the degree of Doctor of Philosophy © Claudine Jacques 2011 ABSTRACT This dissertation is a study of gender transitivity as I see it at work in three of...»





 
<<  HOME   |    CONTACTS
2016 www.book.dislib.info - Free e-library - Books, dissertations, abstract

Materials of this site are available for review, all rights belong to their respective owners.
If you do not agree with the fact that your material is placed on this site, please, email us, we will within 1-2 business days delete him.