Representing Linked Data as Virtual File Systems Bernhard Schandl - - PowerPoint PPT Presentation

representing linked data as virtual file systems
SMART_READER_LITE
LIVE PREVIEW

Representing Linked Data as Virtual File Systems Bernhard Schandl - - PowerPoint PPT Presentation

Representing Linked Data as Virtual File Systems Bernhard Schandl University of Vienna Department of Distributed and Multimedia Systems http://events.linkeddata.org/ldow2009#ldow2009 Madrid, Spain, April 20th 2009 Motivation Linked Data


slide-1
SLIDE 1

Representing Linked Data as Virtual File Systems

Bernhard Schandl University of Vienna Department of Distributed and Multimedia Systems http://events.linkeddata.org/ldow2009#ldow2009 Madrid, Spain, April 20th 2009

slide-2
SLIDE 2

Motivation

  • Linked Data are made for applications to be used.
  • However: desktop applications have no idea about URIs,

HTTP, RDF, Triples, etc.

  • Desktop and Web environments are distinct worlds — we want

to build bridges.

  • Idea: represent Linked Data as virtual file systems.

2

slide-3
SLIDE 3

Potential Applications

  • Unified browsing and navigation of local files and remote LOD

sets.

  • Direct data import and reuse.
  • (Semantic) Integration with desktop resources.
  • and probably some more things we haven’t thought of.

3

slide-4
SLIDE 4

Design Considerations

  • Structural Mismatch: graphs and trees don’t fit easily.
  • Entry Points: where to start?
  • Naming: URIs vs rdfs:labels vs file names
  • Literal values: the data flesh of Linked Data
  • Resource representations: only partially considered in LOD,

but important for desktop applications.

4

slide-5
SLIDE 5

Mapping Approach

  • Virtual directory /!resource/ that holds all resources.
  • Each resource is represented as virtual directory.

5

dbpedia:Berlin dbpedia:Berlin /!resource/ /

slide-6
SLIDE 6

Mapping Approach

  • Object properties are represented as virtual directories

containing symlinks to other resource directories.

6

dbpedia:Berlin dbpedia-owl:City rdf:type dbpedia:Berlin /!resource/ / rdf:type / dbpedia-owl:City /!resource/ / dbpedia-owl:City symlink

slide-7
SLIDE 7

Mapping Approach

  • Datatype properties are represented as virtual files that store

the property’s value.

7

dbpedia:Berlin "891.82"^^xsd:double p:area dbpedia:Berlin /!resource/ / p:area / value-17.txt "891.82"^^xsd:double

slide-8
SLIDE 8

Mapping Approach

  • Incoming object properties are represented as virtual

directories with “is ... of” naming:

8

dbpedia:Berlin dbpedia-owl:City rdf:type dbpedia:Berlin /!resource/ / is rdf:type of / dbpedia-owl:City /!resource/ / dbpedia:Berlin symlink rdf:type dbpedia-owl:City symlink /

slide-9
SLIDE 9

Mapping Approach

  • Resource representations are mapped to virtual files for

different (standard) content types.

9

dbpedia:Berlin dbpedia:Berlin /!resource/ / content-application_rdf.xml dbpedia:Berlin /!resource/ / content-text_rdf.n3 dbpedia:Berlin /!resource/ / dbpedia:Berlin /!resource/ / content-text.csv content-text.html

slide-10
SLIDE 10

Mapping Approach

  • Entry points:
  • Virtual /!resource/ directory:

for known resources.

  • Virtual /<search>/ directory:

for fulltext search results.

10

/ !resource dbpedia:2raumwohnung berlin dbpedia:Berlin p:location rdfs:label dbpedia:Wannsee dbpedia:Vienna dbpedia:Berlin ... dbpedia:Wannsee dbpedia:Berlin value-1.txt value-2.txt ... ... content-text.html content-application_rdf.xml

File Directory Symlink Symbols

slide-11
SLIDE 11

Implementation

  • Prototype available, based on Java and FUSE (sorry Windows

users ...)

  • One SPARQL endpoint is represented as one virtual file

system.

  • Data is retrieved from the endpoint on demand.

11

slide-12
SLIDE 12

Experience: LOD Issues (I)

  • Resource rendering: what is the right way to display a

resource?

  • URI prefixes: how to find them? How to use them in a

consistent manner?

  • Content representation: how do I know which representations
  • f a resource do exist?

12

slide-13
SLIDE 13

Experience: LOD Issues (II)

  • RDF language features: how to deal with collections,

containers, reification, blank nodes?

  • Fulltext queries: regex(), bif:contains, pf:textMatch,

sail:query, ???

  • LOD Updates: would be nice!

13

slide-14
SLIDE 14

14

slide-15
SLIDE 15

15

slide-16
SLIDE 16

16

slide-17
SLIDE 17

17

slide-18
SLIDE 18

18

slide-19
SLIDE 19

19

slide-20
SLIDE 20

Future Work

  • Inclusion of multiple SPARQL endpoints
  • Update (write access) and local modifications
  • Caching

20

slide-21
SLIDE 21

Questions?

bernhard.schandl@univie.ac.at http://www.cs.univie.ac.at/bernhard.schandl http://lodfs.sourceforge.net

21