Semantic Data Integration in myGrid and ourGrid (SEEK)

1
Semantic Data Integration in myGrid and ourGrid
(SEEK)

• National e-Science Centre
• e-Science Institute, Edinburgh
• May 14th, 2004

2
Possible Topics

• Web service description
• Semantics WSDL extensions
• vs standards (driven by own use cases only? wait
  and see?)
• structural/data types
• Pedro files (classify by topic, input,
  properties,)
• what reasoning needed?
• BioMoby work
• Data description (structure and semantics)
• One size fits all? For
• token/packets (data in the small)
• data base schemas (data in the large)
• OGSA-DAI schema format
• role of EML
• Workflow descriptions
• execution models, e.g., Scufl vs Ptolemy
  directors
• data model, data types, explicit type checking
  fa-gtb vs g a-gtb
• semantic types
• Data service glue
• Servicelt-gt serviceShims, ontology-enabled
  conversions

3
Possible Topics

• Use of LSIDs in web services
• a mechanism for pass by reference!?
• our own ad-hoc WSDL extension
• vs. emerging standards
• vs. embedding in a general semantics-enhanced
  WS extension
• GROWL lt-gt Sparrow lt-gt Concrete Abstract Syntax
  lt-gt DIG
• Conversion tools

4
Data description (structure and semantics)

• Mixing physical and logical and ontological in
  existing approaches
• EMBL flat file format, transfer format
• is a standard format -gt captures structure and
  semantics!?
• in ecology no standards (csv,)
• Semantic content description seq-name, seq-id,
  but is NOT formal
• Semantic description is hard to do but not
  impossible
• Taxon transfer schema
• Handling legacy data is harder than w/ new data
• Gene ontology as a successful example
• MGED, MIAME micro array data formats
• Proteomics standard emerging!? Minimum data
• Languages can be mapped BDWorld (eScience)
• Really need transformations tools and a common
  core first
• Biodiversity is behind genomics/proteomics etc
  
• Need for a common semantic annotation/registration
  format?
• Conservation biology example uses biogeography
  and genomics data

5
Data description (structure and semantics)

• Blob data still need to attach semantics but at
  a top-level image data image of reflectance
  measurements?
• Semantic annotation can depend on what is being
  done with the data
• Describe what you can do with a given set of
  (semantic) annotations
• Need for depth of annotation also dependent on
  the use/need for flexibility
• unknown use vs known use vs effort to annote
• ? need to identify formats semantics of the
  whole object semantics of parts within
  (addressing) ability to pull out pieces
• DB to XML where to put the semantics? DB/XML
  doesnt have enough semantics
• Not enough semantics to do what?
• Unless there is some basic agreement on concepts
  eg in OWL, no magic interoperation possible

6
Web service descriptions

• Semantics WSDL extensions
• OWL-S/DAML-S
• DAML-S semantics ws descriptions for
  discovering, enacting, monitoring, composition
  (also has process model), generic,
  domain-independent view
• AI/planning folks support for planning in an
  agent-based framework eg coordination
• Profile discovery process-model
  coordination/WF overlap binding linking
  semantic descriptions to invocation layer (WSDL)
• Downside top-down approach lets develop it
  independently/generic agenty approach quite
  complex if simplifications (form-filling)
  then why use OWL-S?
• Examples amazon shopping, DAML-S virtual
• SEEK/Kepler approach
• Planned Describe inputs and outputs some info
  on the actor semantics
• Ptolemy types
• vs Pedro files schema describes inputs,
  outputs and task semantics
• Holes (vs OWL-S) pre-/post-conditions not there
  (pre good credit post credit decreased by X)
  outside of the WF model

7
Web service descriptions

• WSMO/WSML
• Address specific concern of OWL-S,
• vs standards (driven by own use cases only? wait
  and see?)
• structural/data types
• Pedro files (classify by topic, input,
  properties,)
• what reasoning needed?
• BioMoby work

8
Web service descriptions

• Requirements for semantic annotations for web
  services
• Composability
• browsing for / searching services
• service classification
• Generation of conversion and transformation steps
  (planning)
• DAML-S/OWL-S use in myGrid changes at the
  detailed level? instead of a complex standard
  description what do we need?
• DAML-S process model not used/needed in
  Kepler/Taverna !?
• Profile describes the what
• Process model describes the how
• How does this compare to signature vs. PL
  semantics?
• Level at which DAML-S describes a WF

9
Use of LSIDs in web services

• a mechanism for pass by reference!?
• our own ad-hoc WSDL extension
• vs. emerging standards
• vs. embedding in a general semantics-enhanced
  WS extension
• Why not just a URL? Physical location may change
• Why not URI/URN with namespace?
• LSID gives a resolution protocol
• URNLSIDltresolution-service e.g ku.edugtltscope
  eg GenBankgtidentifierversion
• Resolution sends back WSDL doc to explain whats
  available, e.g. get-data get-meta-data
• Pro standard protocol for accessing the
  metadata use of one protocol across different
  sources
• 2-step resolution get back ws operations then
  apply those on the ids
• Persistence issue resolution service may change
  

10
GROWL lt-gt Sparrow lt-gt Concrete Abstract Syntax
lt-gt DIG

• Conversion tools (format/syntactic level) between
  text-based (and XML-based) formats
• Visualization tools
• Dot,
• GROWL-gt GrOWL
• How many people are actually writing ontologies?
• 1 each -) Chris, Rich,
• Ontology editing tools
• Text-based languages for hackers (power users
  -) GROWL, Sparrow
• CO-ode (ProtégéManchester)
• GrOWL (U Vermont)
• HalfHalf (graphical text-based)
• Intermediate representations to hide logic
  complexity (surgical procedures)
• Inherent problem imprecision of specification
  vs. need for formal
• Very user(-type) dependent as well as type of
  ontology
• Use of Excel spreadsheet (or Word) in practice
• Class hierarchies are easy to get wrong

11
To-do list (2 names each)

• ? DATA need to identify formats semantics of the
  whole object semantics of parts within
  (addressing) ability to pull out pieces
• (Chris, Shawn/Bertram)
• SERVICES SEEK WS describe using Pedro approach
  (vs. Sparrow approach) for semantic WS annotation
  (Chris/Duncun, Shawn,Bertram)
• Workflows Taverna/Scufl lt-gt Kepler/MOML (Ilkay,
  Peter Li)
• ? conservation biology EXAMPLE using biogeography
  and genomics data (Phil, Richard White,
  Matt/Bill? BEAM?)
• ? Use of LSID for SEEK/EcoGrid (Matt, DaveT,
  Martin Senger)
• ? GROWL, Sparrow, etc conversion tools what do
  projects have in terms of such tools joint list?
  E.g., instance store, (Chris, Phil, Shawn)
  Link-Up ontology play group Shared repository
  Reasoning tools