Systems Jan Pettersen Nytun, UiA 1 S O P Following Slides - - - PowerPoint PPT Presentation

1

RDF Storage and Retrieval Systems

Jan Pettersen Nytun, UiA

S O

P

Following Slides - Ref.: Chapter 4 Semantic Web application architecture - Semantic Web for the Working Ontologist, Second Edition: Effective Modeling in RDFS and OWL, May 20, 2011, by Dean Allemang, James Hendler

Jan Pettersen Nytun, UiA, Ontologies, page 2

S O

P

RDF STORE

• An RDF store typically includes a query engine

(SPARQL).

• RDF allows easy merging of data sets (in contrast

to relational data stores).

• RDF stores comes in many flavors: Custom

programmed database solutions, fully supported

• ff-the-shelf products, …

Jan Pettersen Nytun, UiA, page 3

S O

P

Jan Pettersen Nytun, UiA, Ontologies, page 4

Conceptually the simplest relational implementation of a triple store

• A triple store is designed to store and retrieve triple collections of

strings (i.e., Subject-Predicate-Object statements). Each RDF statement is stored as a single row in a three column table.

• Since this fits in a relational database representation, it can be

accessed using conventional relational database tools such as SQL.

S O

P

Jan Pettersen Nytun, UiA, Ontologies, page 5

S O

P

RDF data standards and interoperability of RDF stores

• Relational data stores: Difficult process to

transfer a whole database from one system to another.

• RDF Stores: All have the underlying RDF data

model and support RDF/XML and/or Turtle. Easy to transfer an RDF data set.

Jan Pettersen Nytun, UiA, Ontologies, page 6

S O

P

The common standards also simplifies the issue of federating data that are housed in multiple RDF stores, possibly coming from different vendor sources.

Jan Pettersen Nytun, UiA, Ontologies, page 7

S O

P

RDF query engines

• The SPARQL query language includes a

protocol for communicating queries and results so that a query engine can act as a web service.

• SPARQL endpoints provide access to large

amounts of structured RDF data.

• It is even possible to provide SPARQL access to

databases that are not triple stores, e.g., SPARQL translated to SQL.

Jan Pettersen Nytun, UiA, Ontologies, page 8

S O

P

DATA FEDERATION

• The RDF data model was designed from the

beginning with data federation in mind. Information from any source is converted into a set of triples so that data federation of any kind—spreadsheets and XML, database tables and web pages—is accomplished with a single mechanism.

Jan Pettersen Nytun, UiA, Ontologies, page 9

S O

P

• This extension allows a query author to direct

a portion of a query to a particular SPARQL endpoint.

• Results are returned to the federated query

processor and are combined with results from the rest of the query.

Jan Pettersen Nytun, UiA, Ontologies, page 10

SPARQL 1.1 Federated Query Extension

• SERVICE

S O

P

PREFIX : <http://example/> PREFIX dc: <http://purl.org/dc/elements/1.1/> SELECT ?a FROM <mybooks.rdf> { ?b dc:title ?title . SERVICE <http://sparql.org/books> { ?s dc:title ?title . ?s dc:creator ?a } }

Jan Pettersen Nytun, UiA, Ontologies, page 11

S O

P

RDF Store (Merge)

Jan Pettersen Nytun, UiA, Ontologies, page 12

S O

P

More Details

Following Slides – Ref. article: RDF Storage and Retrieval Systems Alice Hertel, Jeen Broekstra, and Heiner Stuckenschmidt Found in: Steffen Staab - Rudi Studer (Eds.) Handbook on Ontologies Second Edition

Jan Pettersen Nytun, UiA, Ontologies, page 13

S O

P

Possible Generic Architecture of an RDF Store

Jan Pettersen Nytun, UiA, page 14

S O

P

Admin Module

• Gives functionality for adding and deleting

data from the RDF store.

• Loading data from files requires parsing and

validating RDF, consequently an RDF parser and an RDF validator are usually part of the admin module.

Jan Pettersen Nytun, UiA, Ontologies, page 15

S O

P

Query Module and Export Module

• Query module handles queries to the RDF
• store. Implements a parser and handler for
• ne query language.
• Export module allows a dump of the RDF

store into files for data exchange with other systems.

Jan Pettersen Nytun, UiA, Ontologies, page 16

The modules can be accessed locally or remotely, e.g. using SOAP or RMI. This is why the highest layer in the middleware contains protocol handlers that can manage different access modes.

Jan Pettersen Nytun, UiA, Ontologies, page 17

S O

P

Storing RDF data in a relational database requires an appropriate ta

table ble design sign.

There are diffferent approaches that can be classified in generic

neric schemas hemas, i.e.

schemas that do not depend on the

• ntology, and ontology specifi

fic schemas.

Jan Pettersen Nytun, UiA, Ontologies, page 18

S O

P

Simples Generic Schema: One table with three columns named Subject, Predicate and Object

Advantage: No restructuring is required if the

• ntology changes (e.g., new classes, etc., realized

by a simple INSERT command in the table). Disadvantage: Performing a query means searching the whole database and queries involving joins become very expensive.

Jan Pettersen Nytun, UiA, Ontologies, page 19

S O

P

Normalized triple store – a more advanced generic schema

• Requires significantly less storage space.
• One may also split the Triples table into

several tables based on the RDFS properties, e.g., a separate table for RDFS rdfs:subClassOf.

Jan Pettersen Nytun, UiA, Ontologies, page 20

S O

P

Ontology Specific Schemas

• Ontology specific schemas are changing when the
• ntology changes, i.e. when classes or properties

are added or removed.

• The basic schema consists of one table with one

column for the instance ID, one for the class name and one for each property in the ontology. Thus, one row in the table corresponds to one instance.

• Some sort of one-table-per-property and/or one-

table-per class schema.

Jan Pettersen Nytun, UiA, Ontologies, page 21