Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Versioned Triple Pattern Fragments

370 views

Published on

Slides for the presentation at the MEPDaW workshop during ESWC 2017.

Published in: Education
  • Be the first to comment

  • Be the first to like this

Versioned Triple Pattern Fragments

  1. 1. Ruben Taelman - @rubensworks imec - Ghent University Versioned Triple Pattern Fragments A Low-cost Linked Data Interface Feature for Web Archives
  2. 2. Let’s go back in time... 2
  3. 3. Keynote by Axel Polleres - On Archiving Linked and Open Data 30/05/2016 ~ 11:00AM - MEPDaW 2016 3
  4. 4. Keynote by Axel Polleres - On Archiving Linked and Open Data 30/05/2016 ~ 11:00AM - MEPDaW 2016 Importance of dataset evolution, preservation and time-based access 4
  5. 5. Keynote by Axel Polleres - On Archiving Linked and Open Data 30/05/2016 ~ 11:00AM - MEPDaW 2016 Research challenges: … How can certain time-specific queries over archives be answered? ... 5
  6. 6. How can certain time-specific queries over archives be answered at Web scale? 6
  7. 7. Structure of this presentation Requirements for a Web-scale RDF archive interface Interface implementation Demonstration of the interface 7
  8. 8. How can certain time-specific queries over archives be answered at Web scale? 8
  9. 9. How can certain time-specific queries over archives be answered at Web scale? 9
  10. 10. Different ways to query RDF archives Version Materialization (VM) Query a single version Delta Materialization (DM) Query differences between 2 versions Version Query (VQ) Query all versions, results are annotated with their versions Ferná ndez 2016 “Which books were present at time X?” “Which books were added and removed between times X and Y?” “At what times was book A present?” 10
  11. 11. How can certain time-specific queries over archives be answered at Web scale? 11
  12. 12. How can certain time-specific queries over archives be answered at Web scale? VM, DM and VQ queries 12
  13. 13. How can VM, DM and VQ queries over archives be answered at Web scale? 13
  14. 14. How can VM, DM and VQ queries over archives be answered at Web scale? 14
  15. 15. Web scale Allowing a large amount of concurrent clients to query published datasources on the Web with a high availability. 15
  16. 16. Triple Pattern Fragments (TPF) enable SPARQL querying with a low publication cost TPF interfaces enable triple pattern queries to Linked Datasets Both human and machine-readable (hypermedia) Clients evaluate SPARQL queries on their own, using these interfaces → Lowers server load compared to SPARQL endpoints Verborgh 2016 16
  17. 17. How can VM, DM and VQ queries over archives be answered at Web scale? using TPF 17
  18. 18. How can VM, DM and VQ queries over archives be answered using TPF? 18
  19. 19. Some work has been done already into making RDF archives queryable using TPF Query different dataset snapshots with Memento (VM) Continuous queries based on time-annotations (VQ) Vander Sande 2016 Taelman 2016 19
  20. 20. No generic interface that supports all types 20
  21. 21. No generic interface that supports all types yet... 21
  22. 22. VTPF Versioned Triple Pattern Fragments A versioning feature for Triple Pattern Fragments 22
  23. 23. VTPF enables VM, DM and VQ triple pattern queries Additional forms and hypermedia controls. TPF VTPF VM VTPF DM VTPF VQ subject, predicate, object subject, predicate, object, version subject, predicate, object, versionStart, versionEnd subject, predicate, object 23
  24. 24. We introduce the “version” vocabulary for VTPF hypermedia controls and results Each version is seen as a different dataset 24
  25. 25. http://versioned.linkeddatafragments.org/bear 25 Demo time
  26. 26. 26 Human-readable Version Materialization http://versioned.linkeddatafragments.org/bear?subject=http%3A%2F%2Fd.opencalais.com%2Fer%2Fcompany%2Fralg-tr1r%2Fa2472e23-94d8-38e1-9d31-f5d7dd336eaa&predicate=&object=&versionType=VersionMaterialized&version=&versionStart=&versionEnd=
  27. 27. 27 http://versioned.linkeddatafragments.org/bear?subject=http%3A%2F%2Fd.opencalais.com%2Fer%2Fcompany%2Fralg-tr1r%2Fa2472e23-94d8-38e1-9d31-f5d7dd336eaa&predicate=&object=&versionType=DeltaMaterialized&version=&versionStart=&versionEnd= Human-readable Delta Materialization
  28. 28. 28 http://versioned.linkeddatafragments.org/bear?subject=http%3A%2F%2Fd.opencalais.com%2Fer%2Fcompany%2Fralg-tr1r%2Fa2472e23-94d8-38e1-9d31-f5d7dd336eaa&predicate=&object=&versionType=Version&version=&versionStart=&versionEnd= Human-readable Version Query
  29. 29. 29 <http://versioned.linkeddatafragments.org/bear#dataset> hydra:search _:patternVersionMaterialized. _:patternVersionMaterialized hydra:template "http://versioned.linkeddatafragments.org/bear?versionType= VersionMaterialize d{&subject,predicate,object,version}"; hydra:variableRepresentation hydra:ExplicitRepresentation; hydra:mapping _:subject, _:predicate, _:object. _:subject hydra:variable " subject"; hydra:property rdf:subject. _:predicate hydra:variable " predicate"; hydra:property rdf:predicate. _:object hydra:variable " object"; hydra:property rdf:object. _:patternVersionMaterialized hydra:mapping _:version. _:version hydra:variable " version"; hydra:property ver:relatedVersion. Machine-readable hypermedia controls
  30. 30. 30 <http://versioned.linkeddatafragments.org/bear?versionType=VersionMaterialized&version=8 > a ver:Version. Each version is a regular TPF entrypoint http://client.linkeddatafragments.org/#datasources=http%3A%2F%2Fversioned.linkeddatafragments.org%2Fbear%3FversionType%3DVersionMaterialized%26version%3D8&query=CONSTRUCT%20WHERE%20%7B%0A%20%20%3Fs%20%3Fp%20%3Fo.%0A% 7D%0ALIMIT%20100 So you can query against all versions using a TPF-enabled client
  31. 31. But we’re not done yet! We considered the interface for querying RDF archives We still need mapping to RDF archive query language(s) can we come up with a standard? appropriate back-end storage solutions snapshot, change-based, time-based, hybrid? 31

×