Semantically Resolving Type Mismatches in Scientific Workflows

1
Semantically Resolving Type Mismatches in
Scientific Workflows

• Derouiche Kheiredine
• kd05r_at_ecs.soton.ac.uk
• School of Electronics and Computer Science
• University of Southampton
• November 2007

2
Scientific Workflows
Semantically Resolving Type Mismatches in
Scientific Workflows

• Scientific workflows describe structured
  activities arising in scientific problem-solving.
• Conducting experiments involve complex and
  structured computations.
• Semantic mismatches among resources involve much
  human intervention.
• Participating services are owned by different
  organizations, defining compensations is critical
  to a successful recovery from failures.

3
Workflows in Bioinformatics

• Integrating different tools to solve biological
  problems
• Usually involves
• Manual data transfer between applications
• Understanding data formats
• Converting file formats where appropriate
• Manual workflows involve a large number of steps.
  Manual execution is time-consuming and
  error-prone
• User is required to possess a deep knowledge and
  understanding of disparate application
  environments

4
Using tools in Bioinformatics
Specific Services
Task
Service Class
WSWUBlast Service Operation
Blast Service
DNA Sequence Similarity Search
blastn (query, database, email)
Specification of an in silico experimental
design Sequence Similarity Search
5
Automated Workflows

• Make the task of creating a workflow a simple
  drag and drop process
• Make the resulting workflow diagram self
  documenting, showing exactly how to perform
  bioinformatics experiment
• Automatic execution of steps specified in
  workflow
• Monitoring workflow execution to help debugging
  and intervention
• Reduces complexity for scientific users, as well
  as support sharing and allow repeatability

6
Bioinformatics Workflow Systems

• Specialized workflow systems designed to develop
  workflows in bioinformatics
• Different workflow standards and systems
• BPEL Business workflow standard adapted for
  scientific workflows
• UNICORE a Grid middleware, it provides a GUI
  for workflow devcelopment
• Globus an open source toolkit implementing many
  Grid related standards
• Kepler graph based modelling language to develop
  workflows
• Taverna Workbench choreography tool for
  bioinformatics Web Services
• Triana develop component based workflow and
  provide coupling with Grid middleware tools

7
Windows Workflow Foundation (1)
Semantically Resolving Type Mismatches in
Scientific Workflows

• Part of .NET Framework 3.0
• Workflows are a collection of activities.
• Components
• Base Activity Library Out-of-box activities and
  base for custom activities.
• Runtime Engine Workflow execution and state
  management.
• Runtime Services e.g. RDBMS, persistence,
  transactions
• Visual Designer Graphical and code-based
  construction in Visual Studio or standalone

8
Windows Workflow Foundation (2)
Semantically Resolving Type Mismatches in
Scientific Workflows
9
Semantic Web Services
Semantically Resolving Type Mismatches in
Scientific Workflows

• The augmentation of Web service descriptions with
  Semantic annotations.
• Aims to automate Web service discovery,
  composition, invocation, and monitoring.
• Two different approaches
• Revolutionary OWL-S, and WSMO.
• Evolutionary WSDL-S, and SAWSDL.
• The SAWSDL approach builds on existing Web
  service standards and is agnostic to ontology
  representation.

10
Semantic Annotations for Web Services Description
Language
Semantically Resolving Type Mismatches in
Scientific Workflows

• SAWDL is an extension of WSDL using the
  extensibility elements.
• Two basic types of annotations
• Model reference, associates selected WSDL
  components with Semantic concepts.
• Schema mapping, deals with data heterogeneity by
  transforming one data representation into
  another.
• Annotations for WSDL 1.1 and WSDL 2.0.
• API and tool support including SWASDL4J,
  Woden4SAWSDL, Radiant...

11
SAWSDL Scope
Semantically Resolving Type Mismatches in
Scientific Workflows
Annotated using modelReference Annotated using
modelReference with schemaMapping
Note - All elements may have ltdocumentationgt as
first child
12
SAWSDL Example
Semantically Resolving Type Mismatches in
Scientific Workflows

• ltwsdldefinitions targetNamespace"http//www.w3.o
  rg/2002/ws/sawsdl/spec/wsdl/order"
• xmlnswsdl"http//schemas.xmlsoap.org/wsdl/"
• xmlnsxs"http//www.w3.org/2001/XMLSchema"
• xmlnssawsdl"http//www.w3.org/ns/sawsdl"gt
• ltwsdltypesgt
• ltxselement namepurchaseOrderResponse
  typexsstring
• sawsdlmodelReference"http//www.w3.org/2002/ws/s
  awsdl/spec/ontology/purchaseorderPurchaseOrderRes
  ponse"
• sawsdlliftingSchemaMapping"http//www.w3.org/200
  2/ws/sawsdl/spec/mapping/Response2Ont.xslt"gt
• lt/xselementgt
• lt/wsdltypesgt
• ltwsdlportType namePurchaseOrder"gt
• ltwsdloperation name"order"gt
• ltsawsdlattrExtensions
  sawsdlmodelReference"http//www.w3.org/2002/ws/
  sawsdl/spec/ontology/purchaseorderRequestPurchase
  Order"/gt
• ltwsdlinput
• messageLabel"OrderRequestMessage"
• elementpurchaseOrderRequest"/gt
• ltwsdloutput
• messageLabel"OrderResponseMessage"

13
Leveraging existing Java for .NET
Semantically Resolving Type Mismatches in
Scientific Workflows

• A C implementation of the SAWSDL specification.
• Support of Model Reference annotations, OWL/RDF
  definitions.
• Lifting/Lowering schema support, XSLT/SPARQL
  mapping definitions.
• Allows the creation of SAWSDL based applications.
• Extends the .NET API for WSDL1.1.
• Support for WSDL2.0 through XSLT.

14
Implementation
Semantically Resolving Type Mismatches in
Scientific Workflows

• Development of a custom activity that extends the
  base Web Service activity shipped with WF.
• Enables a semi-automatic composition of Semantic
  Web Services, and the execution of the workflow.
• Can be composed with Web Services described using
  WSDL files.
• A C implementation of the activity.
• Semantic capabilities are provided by the Jena
  library, integration with C is enabled via IKVM.

15
Semantic Reasoning
Semantically Resolving Type Mismatches in
Scientific Workflows

• Model reference annotations describe the
  functionalities a Web service provides.
• Use ontologies as semantic models for the
  semantic annotations.
• Reasoning capabilities are provided by using
• Jena, an open source Semantic Web framework for
  Java.
• Pellet, an open source Java OWL-DL reasoner.
• Currently support schema type and message part
  annotations to achieve automatic parameter
  binding.

16
Schema Type Mapping
Semantically Resolving Type Mismatches in
Scientific Workflows

• Provide mappings between XML and semantic models.
• Lifting Schema Mapping specifies mapping between
  WSDL Type Definitions in XML and semantic data.
• Used XSTL and XQuery as mapping languages.
• Lowering Schema Mapping specifies mapping between
  semantic data and WSDL Type Definitions in XML.
• Used SPARQL to query ontology, followed by XSTL
  and XQuery.
• Semantic data is queried through SPARQL, it is
  supported by Jena through its query engine.

17
IKVM .NET
Semantically Resolving Type Mismatches in
Scientific Workflows

• An implementation of Java for the Microsoft .NET
  Framework.
• It includes the following components
• A Java Virtual Machine implemented in .NET.
• A .NET implementation of the Java class
  libraries.
• Tools that enable Java and .NET interoperability.
• Used to compile Jena and Pellet JAR libraries
  into .NET DLL assemblies, Java bytecode is
  translated to Common Intermediate Language (CIL).
• Allowed using Jenas capabilities in the
  implementation of the Semantic Web service
  activity.

18
Semantic Web Service Activity (1)
Semantically Resolving Type Mismatches in
Scientific Workflows

• Activity bindings are the key feature that
  enables property binding between activities, or
  on the workflow itself.
• This mechanism allows data propagation between
  composed activities.
• WF rely on syntactic approaches when binding
  properties between activities.
• The SWS activity implements a basic semantic
  matching engine to better support semantically
  compatible properties.

19
Semantic Web Service Activity (2)
Semantically Resolving Type Mismatches in
Scientific Workflows

• Automatically bind SWS parameters to composed
  workflow activities using the semantic approach.
• The semantic model annotation of an activitys
  input has to be equivalent or a subclass of the
  composed activitys output one.
• Values are mapped to the appropriate data
  representation at design time.
• Missing activity bindings can be manually added
  using the WF visual designer.

20
Bioinformatics Workflow Example
accession
X14298
GetEntry
getFASTA_DDBJEntry
database
email
sequence
Atgagtgatggagcagttcaaccagacggtggtcaacctgctgtcagaaa
tgaaagagctcaggatctgggaacgggtctggaggcggg
embl
kd05r_at_ecs.soton.ac.uk
WSWUBlast
blastn
jobID
M7WEXBN7013
21
Automatic Binding in Bioinformatics Workflow
GetEntry
getFASTA_DDBJEntry
SAWSDL
DNA Sequence
Sequence
Semantic Concept
Output
Bind Parameters, Carry out necessary translations
Semantic Reasoner
Input
DNA Sequence
Sequence
SAWSDL
WSWUBlast

• Degrees of Match
• Exact
• Subclass

blastn
22
Conclusion Future Work
Semantically Resolving Type Mismatches in
Scientific Workflows

• API implementations that enable the development
  of semantically annotated Web services.
• Semantic Web service activity integration to WF,
  facilitating workflow building and manipulation.
• Future Work
• Improve the SWS activity by processing more
  SAWSDL annotations, e.g. operation and portType.
• Semantically annotate Bioinformatics Web
  services, then use WF to build a workflow
  composed of SWS activities in order to test the
  implementation.
• Implement an approach to semantically guide and
  verify compensations and exceptions.