Page 1
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance
Robert Sanderson
[email protected]
The Provenance of Madame Bonnier:Museum Linked Data
With Thanks To:
Ruth CuadraBrenda PodemskiDavid Newbury
Kelly DavisJoan Cobb
Vladimir Alexiev…
Page 2
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance • Case Study: Madame Bonnier
• Data Models: Integration of Object and Provenance• Lessons Learnt• Data Ownership vs Stewardship• Data Quality• Model Alignment• Structure or Interaction?
Overview
Page 3
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance Portrait of Madame Bonnier de
la Mosson as DianaPaintingJean-Marc Nattier (1685 – 1766)France, 1742Oil on Canvas129.5 x 96.8 cm77.PA.87Getty MuseumOn View: South Pavilion, S202
Madame Bonnier
Page 4
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance
Madame Bonnier: Provenance
Page 5
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance
Madame Bonnier: Provenance
Museum Provenance Index
Page 6
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance
Madame Bonnier: Provenance?
Provenance Index Provenance Index
Page 7
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance
Madame Bonnier: Provenance?!
Page 8
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
anceI’m Thinking …
Page 9
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
anceBut Actually …
Page 10
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance
Error in the Data :(
Page 11
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
anceEveryone As
Diana At Some Point
Page 12
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
anceEveryone As
Diana At Some Point
Page 13
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
anceEveryone As
Diana At Some Point
Ran out of red paint?
Page 14
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance Portrait of Madame Bonnier de
la Mosson as DianaPaintingJean-Marc Nattier (1685 – 1766)France, 1742Oil on Canvas129.5 x 96.8 cm77.PA.87Getty MuseumOn View: South Pavilion, S202
Madame Bonnier
Page 15
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance
Museum Data Model: Description
Page 16
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance
Museum Data Model: Provenance
Page 17
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance
Museum Data Model: Complete
Page 18
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance
Object Data Model: Complete
Start of Provenance
Current Provenance
Page 19
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance
General Provenance Model
Page 20
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance
Auctions Model
Page 21
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance
Lessons Learnt
Page 22
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance
Data Ownership vs StewardshipProvenance Index makes assertions about others’
objects (including Museum)
Page 23
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance
Data Ownership vs StewardshipProvenance Index makes assertions about others’
objects (including Museum)
Authoritative about the documents … … but not about the objects’ provenance
Page 24
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance
Data Ownership vs StewardshipProvenance Index makes assertions about others’
objects (including Museum)
Authoritative about the documents … … but not about the objects’ provenance
=> Acting as a steward for the provenance, not the owner of it
Page 25
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
anceData Quality
Linked Data makes errors easier to find … … but doesn’t correct them for you
Page 26
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
anceData Quality
Linked Data makes errors easier to find … … but doesn’t correct them for you
Quality of LOD is limited by quality of non LOD input, for purpose of linking
Page 27
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
anceData Quality
Linked Data makes errors easier to find … … but doesn’t correct them for you
Quality of LOD is limited by quality of non LOD input, for purpose of linking
Beware false precision=> Automatic inferencing is risky
Page 28
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance
Model AlignmentAligning models is hard, while avoiding false
precision!
Page 29
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance
Model AlignmentAligning models is hard, while avoiding false
precision!
Need to get over “need” for 100% precision, 100% coverage, 100% completeness
Page 30
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance
Model AlignmentAligning models is hard, while avoiding false
precision!
Need to get over “need” for 100% precision, 100% coverage, 100% completeness
=> Start somewhere, then implement, learn, correct, and iterate
Page 31
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance
Structure or InteractionLinked Data as a Data Model … provides single structure to implement, query … acceptable if complete and accurate
Linked Data as an API … provides data to integrate, use … acceptable if understandable and usable
Page 32
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance
Structure or InteractionLinked Data as a Data Model … provides single structure to implement, query … acceptable if complete and accurate
Linked Data as an API … provides data to integrate, use … acceptable if understandable and usable
=> As completeness increases, usability decreases
Page 33
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance
If it’s not usable, why are we doing it?
Page 34
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
anceFinal Thoughts
Need to focus on our goals … not on religious wars around ontologies
Need to find the balance between accuracy and usability
Need to recognize this is HARD and we’re not going to get it right the first time. Or the second.
Need to keep on keeping on :)
Page 35
@azaroth42
[email protected]
LOD
Less
ons L
earn
t: Pr
oven
ance
Rob Sanderson / [email protected]
Thank You!