SeaDataNet Ontology Use Case

1
SeaDataNet Ontology Use Case
Coastal Atlas Interoperability Workshop,
Corvallis, July 17-19 2007
( Lessons Learned)

• Roy Lowry
• British Oceanographic Data Centre

2
Summary

• What is SeaDataNet?
• Some SeaDataNet semantic issues
• What has SeaDataNet done?
• What is SeaDataNet going to do?
• Is SeaDataNet relevant to CAI?

3
What is SeaDataNet?

• SeaDataNet in a Nutshell
• Combine over 40 oceanographic data centres across
  Europe into a single interoperable data system
• Approach is to adopt established standards and
  technologies wherever possible
• Two phases
• One brings 12 centres together with centralised
  metadata and distributed data as files. Due fully
  operational in autumn 2008 (beta next February)
• Two introduces data virtualisation, aggregation,
  cutting and 30 more centres. Due in 2010
• Project is well on its way up the
  interoperability operational implementation curve

4
SeaDataNet Semantic Issues

• The major problem facing the project is
  heterogeneous legacy content
• SeaDataNet inherited 3 independently-developed
  metadatabases
• Each is heavily populated (3000-30000 records)
• Each had its own independently developed
  controlled vocabularies
• These vocabularies
• Covered overlapping domains
• Said similar things in different ways
• Provided a shining example of how NOT to manage
  vocabularies

5
Brief Diversion

• Vocabularies can have two types of governance
• Content governance
• Mechanism for making decisions on vocabulary
  population
• Expected deliverables include
• Vocabulary standards and internal consistency
• Change on a timescale matching the needs of the
  user community
• Terms with definitions!!!
• Technical governance
• Vocabulary storage, maintenance and serving
• Expected deliverables include
• Convenient access to up to date vocabularies
• Clear, rigorous vocabulary versioning
• Version history through audit trails
• Maintenance that doesnt break user systems

6
SeaDataNet Semantic Issues

• Vocabulary content governance
• Done by individuals who were often inadequately
  qualified to do the job
• Metadata entry form with an Add to Vocabulary
  button used by students
• Vocabulary technical governance
• Scattered files on servers or inaccessible
  database tables
• Multiple data models (e.g. some with
  abbreviations, some without)
• No versioning
• Vocabularies updated by destructive overwrites
• Harmonisation required for related vocabularies
• Within centralised metadata
• Between partner local systems and centralised
  metadata

7
What has SeaDataNet Done?

• Established content governance
• Within SeaDataNet (TTT e-mail list)
• Further afield (SeaVoX e-mail list)
• Established technical governance
• Adopted the NERC DataGrid Vocabulary Server
• Heavily defended Oracle back end
• Automated version and audit trail management
• Web Service API front end plus clients e.g.
  http//vocab.ndg.nerc.ac.uk/client/vocabServer.jsp
  
• Currently serving out 75 lists
• Established a Mapping Infrastructure
• List entries connected by SKOS RDF triples
• Operational mappings between parameter
  vocabularies (GCMD science keywords, CF Standard
  Names)

8
What is SeaDataNet Going To Do?

• Harmonise centralised metadata vocabularies or
  map if too hard
• Map centralised vocabularies to partner system
  vocabularies
• Build metadata crosswalks and generators (e.g.
  from CF) that include semantics (Use case 1)
• Implement Smart Discovery for legacy plaintext.
  E,g. search for pigment, find chlorophyll (Use
  case 2)
• Establish URLs to represent vocabularies and
  individual entries delivering XML probably SKOS
  documents
• Extend mapping efforts to other areas such as
  devices
• Release a much improved Vocabulary Server API
  (mid-August)

9
Is SeaDataNet Relevant to CAI?

• This workshop is about building a coastal atlas
  ontology that brings together semantic resources
  that say similar things in different ways
• The vocabulary entry semantic content may be
  different from oceanographic parameters, but the
  problem is essentially the same
• If it works for SeaDataNet it will probably work
  for the CAI community
• More important if it didnt work for SeaDataNet
  then it probably wont work for CAI

10
Is SeaDataNet Relevant to CAI?

• What has worked for SeaDataNet
• The NERC DataGrid Vocabulary Server
• Content governance through a MODERATED e-mail
  list (also works pretty well for CF Standard
  Names)
• Representing vocabulary terms by URNs in metadata
  documents
• What I believe will work in the next 12 months
• Semantic interoperability through mappings
• The conceptual framework of RDF in general and
  SKOS in particular
• 21st Century tooling

11
Is SeaDataNet Relevant to CAI?

• What hasnt worked for SeaDataNet
• Weak content governance
• Examples
• Terms without definitions
• Vocabularies without strict entity definitions
  populated by mixed entities e.g.
• helicopter class
• RRS Discovery instance
• Vocabularies without managed deprecation
• Poor technical governance
• Example
• A vocabulary served by
• Dynamic web page from database
• Static HTML page
• ASCII file as e-mail attachment
• Each having a different number of entries.

12
Thats All Folks!

• Thank you for your attention
• Any questions?
• Morals
• Always provide definitions for your terms
• If you are going to use vocabularies to build an
  ontology make sure that they are properly governed