Mapping Museum Metadata to the CRM in Sculpteur

1
Mapping Museum Metadata to the CRM in Sculpteur

• Presented By Patrick Sinclair
• IAM, University of Southampton,
  UKpass_at_ecs.soton.ac.uk

2
Contents

• Sculpteur Overview
• The Semantic Layer
• Based on the CRM ontology
• Our experiences with the CRM within Sculpteur
• Navigating the semantic layer

3
SCULPTEUR
3 Millon Euros, 3 year project (May 2002 May
2005)
Uffizi Gallery
4
Problem Domain
5
Existing Systems
6
Overview of Sculpteur

• Integrated concept, metadata and content based
  browsing and retrieval of museum information
• Use of semantic web technologies will enable
• Enhanced navigation and exploration
• Automatic data augmentation
• Flexible systems integration and interoperability
• Content based retrieval
• Image content analysis
• 3D object analysis
• E-learning

7
Searching by content

• Find objects that have a pattern / colour similar
  to this image
• Find vases that are a similar shape to this one
• Find paintings with cracks like this
• What type of mould was used to make this figurine?

8
The Semantic Layer

• Motivation
• The mapping process
• The concept browser

9
Museums are rich in information

• Museums already have information about
• Works of art (title, medium, state of
  restoration)
• Creators of works of art (name, date of birth)
• Dates (artist date of birth, date of photograph)
• Locations (where a work of art is stored,
  artists country of origin)
• Digital representations (images, models, angle of
  lighting, full or sub image)
• Museums are creating large numbers of 2D images
  and 3D models and movies

10
Questions using concepts

• Which artists have lived, worked or were born in
  Normandy?
• Which countries do I have 18th century paintings
  from?
• What is the oldest Polynesian wooden object?
• Find images of the reverse of paintings which
  have undergone restoration

11
Navigating the Semantic Layer
12
Data Augmentation

• Information about museum collections is sometimes
  incomplete e.g.
• Dates when paintings were created
• Places where artists where born
• Augmentation agents will automatically obtain
  missing information from the semantic web
• Not much semantic web around!
• Initial agent uses natural language processing on
  unstructured web pages

13
Benefits

• New ways to search
• Scientists, Public, Curators, Teachers
• Graphical presentation
• Searching by example
• Easy to use
• Sharing of information between museums
• Searching collections owned by different
  organisations
• Support for multiple languages
• Users can still use traditional and familiar
  approaches

14
First Prototype
15
CIDOC Conceptual Reference Model (CRM)

• Ontology for documenting artefacts within the
  museum and gallery domain
• 81 classes and 139 properties
• 10 years of development
• Draft ISO standard

16
CIDOC CRM in Sculpteur

• Comprehensive description of art domain
• Model based on formal principles of ontology
• Established and constantly being revised
• Enhances communication performance with partners
• Multilingual functionality possible

17
Ontology Mapping
CIDOC Conceptual Reference Model
Knowledge Base
Instances
Actors
Events
Objects
Background knowledge / Authorities
Derived knowledge
Sources and metadata
Source ICS-Forth
18
Mapping Subset

• Museums identify areas of interest
• Subset of CRM ontology provided to each museum
  partner depending on the concepts present in the
  legacy metadata

19
Mapping in Sculpteur

• Held an ontology workshop for museum partners
• Determine ontological scopes
• Select domains of interest
• Map metadata to CRM
• Museums had trouble performing mapping
• Technical partners attempted to create initial
  mapping
• Contacted a CRM expert Patrick Le Boeuf

20
Mapping Example
Table Field Type
Local art object metadata Width Integer
DB Schema
Master Table Primary key Connection key to secondary table
Art object Art object id Local object id
CRM
Domain Link Range
E84.Information Carrier P43F.has dimension E54.Dimension
E54.Dimension P2F.has type E55.Type
E54.Dimension P91F.has unit E58.Measurement Unit
E54.Dimension P90F.has value E60.Number

• Tool generates subset of CRM ontology from
  mapping table

21
Extending the CRM

• Different degree of specificity between CRM and
  museum datasets
• Extended E.55 Type to model activities not
  covered by CRM
• Added concepts and relationships
• E.g. painting, has_related_artists

22
Extending the CRM (cont.)

• Handling mapping of multiple metadata fields to
  one CRM concept
• Modifications performed in Protégé
• Ontology represented in RDF Schema

Metadata Domain Link Range
MarksInscriptionsAuthor E84.Information_Carrier P65F.shows_visual_item E37.Mark
E37.Mark P105F.right_held_by E39.Actor
E39.Actor P131F.is_identified_by E82.Actor_Appellation
AssociatedName E84.Information_Carrier P62F.depicts E7.Activity
E7.Activity P11F.had_participant E39.Actor
E39.Actor P131F.is_identified_by E82.Actor_Appellation
23
Populating Concept Browser

• Tools for populating concept browser with
  instance information
• Collect instances from museum databases
• Structure information according to ontology
• Script is used to generate instances from museum
  database
• Instances stored as RDF
• Generated at installation
• Used within Concept Browser

24
Mapping in Sculpteur

• Complex process!
• Close collaboration between museums, technical
  partners and external experts.
• Expert assistance crucial for creating and
  validating mappings.
• Iterative and involved
• Collaboration has been time and effort consuming,
  but essential for achieving an accurate and
  meaningful mapping

25
Mapping problems

• Getting to grips with the CRM
• Understanding the legacy data
• Mapping the process of photographing objects
• Mapping is complex for museum partners - simple
  fields map to long chains
• E.g. School name

E84 Information Carrier P108B was produced by E12 Production Event
E12 Production Event P14 carried out by E21 Person
E21 Person P107B is current or former member of E74 Group
E74 Group P2 has type E55 Type (value "school")
E74 Group P131 is identified by E82 Actor Appellation
26
Mapping Problems (cont.)

• Complex to determine relevant mapping
• Many of the fields in the database are empty
• Several fields can contain similar information,
  but from different perspectives
• Inconsistencies in legacy information missing
  entries, merged fields, misspelling, excessive
  creativity of database maintainer

27
Unresolved issues

• Merged fields
• e.g. materials field can also describes creation
  event
• Some metadata fields contain complex/non-atomic
  values that express relationships between records
• Data not handled in CRM
• e.g. genealogical data in Uffizi database

28
Generating Instances Problems

• Each museum has own metadata legacy system and
  format
• Imported into Sculpteur mySQL relational database
• Metadata delivered in various formats (XML, CSV,
  database dumps) and requires pre-processing
• Includes format transformation and
  cleaning/consistency checking
• Manual process that is then encapsulated within a
  software tool (customised for each partner) so
  process can be repeated against larger datasets
• Tools not ready to be deployed at user sites, so
  requires metadata and mappings in advance

29
Concept Browser

• Show the ontology information graphically
• Support novice users browsing domain
• Extending TouchGraph
• Open source
• Provides a dynamic graph layout interface
• Jena toolkit for ontology and instance information

30
Concept Browser (cont.)

• Concept Browser
• Show associated relations when a user selects a
  specific concept
• Collapse concepts in the sub/super hierarchies
• Explicitly display sub/super relations between
  relations
• Allow users to view instances for a given class
• Display control lists as a dynamic graph

31
Navigation Facilities

• Provide navigation tools to help the user browse
  the ontology
• Buttons
• Who - E39.Actor
• What - E18.Physical Stuff
• When - E50.Date and E52.Time-span
• Where - E53.Place
• How - E7.Activity
• List of Concept Names
• Tree View

32
Property View
33
Querying for Instances
34
Controlled Lists
35
Concept browser problems

• Extending Touchgraph
• e.g. properties require overlapping edges
• Usability
• First evaluation users had trouble
• RDF slow to load, too many instances!
• First prototype
• Only basic information stored
• Unconnected to images and other rich information
  in legacy system

36
Control List Problems

• Control lists contain inconsistent values
• E.g. top and toop
• Ideally chosen by museum collaboration required
  to properly understand semantics of the data
• Inferring may result in new instances not in
  legacy database
• May require further inference
• e.g. use a geographical ontology for location
• Process difficult to automate dependent on
  structure and semantics of metadata, and these
  semantics are not explicit in the data

37
Concept Browser next steps

• Evaluation results
• Ontology simplification
• Ontology shortcuts
• Interested in suggestions!

38
Demo Videos
39
More Information

• Try Sculpteur yourself!
• http//piltdownman.it-innovation.soton.ac.uk/
• Sculpteur User Interest Group
• http//www.sculpteurweb.org/html/sig.htm