WS2: Work Package 2

Annotated Service Invocation

Execute (services described by) WSDL interfaces to answer user queries.

Procedure

• Identify the information in web services.
  • annotate existing descriptions
• Query that specific information.

  • SELECT ?asin ?title WHERE {
    _:b tns:id "0FWYBWB91M5S26YBE382" ;
        tns:keywords "Weaving" ;
        tns:index "Books" ;
        tns:asin ?asin ;
        tns:title ?title ;
        tns:page ?page }

• Automatically invoke the appropriate service(s).

What's New?

• Modeled operations' inputs and outputs as rules.
• Modeled parameters in user query as ground facts.
• Choreography!
• Separate documentation for SPARQL Anntoations and SPARQL Annotations on WSDL.

Features of SPARQL in WSDL (SPDL)

• Annotations on web service operations
• Builds RDF graph query pattern
• Uses (relatively familiar) SPARQL query pattern
• — Sufficient for invocation given a grounding for the pattern

Connecting SPARQL to XML

•   <ItemAttributes>
       <Title>Learning to Weave, Revised Edition</Title>   
       <Author>Deborah Chandler</Author>                   
       <Author>Debbie Redding</Author>                     
    </ItemAttributes>

•   <xs:element name="Item">
      <xs:complexType spat:SPAT='?book tns:doctitle xpath("aws:Title")...'>
        <xs:sequence>
  	<xs:element name="Title" type="xs:string" minOccurs="0"/>
  	<xs:element name="Author" type="xs:string" minOccurs="0" maxOccurs="unbounded"/>
        </xs:sequence>
      </xs:complexType>
    </xs:element>

• 

Identifying information

• Using SPARQL Annotations (SPARQL + XPath) in WSDL:
  • SPARQL: SQL-like RDF query language

    SELECT ?tile WHERE { ?response tnss:doctitle ?title }

  • XPath: identify locations in a document

    aws:ItemSearchResponse/aws:Items/aws:Item/aws:Title

  • WSDL/XSD: associate messages (documents) with services

    <definitions><types><element name="ItemSearchResponse"></element></types>...
        <message name="ItemSearchResponseMsg"/></definitions>

• SPARQL annotations tie RDF data to locations in a Web Services message.

  • <xs:element name="Item" spat:SPAT='?book tns:doctitle xpath("aws:Title")'/>
      

Demo excerpts

PREFIX tnss: <https://webservices.amazon.com/AWSECommerceService/2004-11-10>
SELECT ?title ?page WHERE {
_:b tnss:id "0FWYBWB91M5S26YBE382" ;
    tnss:keywords "Weaving" ;
    tnss:index "Books" ;
    tnss:asin ?asin ;
    tnss:doctitle ?title ;
    tns:page ?page }

SAWSDL Specification

• Semantic Annotations for WSDL and XML Schema
  • W3C Candidate Recommendation
  • Defines:
    • modelReference: association between a WSDL entity and a concept in some semantic model
      • XSD complex types, simple types, elements, and attributes
      • WSDL operations and interfaces
    • liftingSchemaMapping and loweringSchemaMapping;, transformation of information from Web Service messages to semantic model
      • XSD elements, complex types and simple types

SAWSDL Specification

• Accompanying Usage Guide references:
  • SPARQL Anntoations
  • SPARQL Annotations on WSDL
  and describes their use in SAWSDL.

SPDL Expressed in SAWSDL

a sawsdl:loweringSchemaMapping:

      <xs:element name="Item">
        <!-- xs:complexType spat:SPAT='?book tns:doctitle xpath("aws:Title") -->
        <xs:complexType sawsdl:loweringSchemaMapping='&AWS;title'>
	  <xs:sequence>
	    <xs:element name="Author" type="xs:string" minOccurs="0" maxOccurs="unbounded"/>
	    <xs:element name="Title" type="xs:string" minOccurs="0"/>
	  </xs:sequence>
	</xs:complexType>
      </xs:element>

points to a document containing the SPDL annotations:

:id       spat:SPAT '?req :id xpath(".")' .
:keywords spat:SPAT '?req :keywords xpath("tns:Keywords") ; :index    xpath("tns:SearchIndex")' .
:asin     spat:SPAT '?book :asin xpath("tns:ASIN")' .
:title    spat:SPAT '?book tns:docTitle xpath("aws:Title")' .

Operation Rules

• Model information in Input and Output as a graph pattern.
• Allow engine to infer service invocation paths.

Very simple inference appears to model reasonable use cases.

Rule Construction Details

• Traverse operation input and output:

      <operation name="ItemSearch">
        <input message="tns:ItemSearchRequestMsg"/>
        <output message="tns:ItemSearchResponseMsg"/>
      </operation>
  

  for the annotations:

      <xs:element name="ItemSearch"
       spat:SPAT='?req tns:asin xpath("tns:ASIN")'>
          ...
            <xs:element name="ASIN" type="xs:string" minOccurs="0"/>
          ...
      </xs:element>

• Build rule from rule body (Input) and rule head (Output):

  ┌ op {https://webservices.amazon.com/AWSECommerceService/2004-11-10}ItemLookup: 
  │     fwrule ask(?req this:id XPath(., 91642dc) && 
  │                ?req this:asin XPath(., 919d638))
  │            assert(?req this:page _:<>:8.XPath(this:Items/this:Item/this:DetailPageURL, 91d37c8) && 
  │                   ?req this:reqId _:<>:9.XPath(., 9211c38) && 
  │                   ?req this:asin _:<>:10.XPath(tns:ASIN, 920b2b0) && 
  │                   ?req this:title _:<>:11.XPath(tns:Title, 9298a50))
  │               -->{<qname:{https://webservices.amazon.com/AWSECommerceService/2004-11-10}ItemLookup>}

Rule Invocation Details

The ground facts from the users query:

PREFIX tns: <https://dev.w3.org/cvsweb/perl/modules/W3C/SPDL/descriptions/AWSECommerceService-SAWSDL-item-terms.ttl#>
SELECT ?title ?page WHERE {
_:b tns:id "0FWYBWB91M5S26YBE382" ;
    tns:keywords "Weaving" ;
    tns:index "Books" ... }

combined with a rule head:

┌ op {AWS:ItemSearch: 
│     fwrule ask(?req this:id XPath(., 8ae97f0) && 
│                ?req this:keywords XPath(tns:Keywords, 91b1d40) && 
│                ?req this:index XPath(tns:SearchIndex, 91b1d40))
│            assert(?req this:reqId _:19.XPath(., 92b17b0) && 
│                   ?req this:asin _:20.XPath(tns:ASIN, 92d6890) && 
│                   ?req this:title _:21.XPath(tns:Title, 93150e0))
│               -->{AWS:ItemSearch}

produce new inferences:

_:21 this:reqId _:19.XPath(., 92b17b0) .
_:21 this:asin _:20.XPath(tns:ASIN, 92d6890) .
_:21 this:title _:21.XPath(tns:Title, 93150e0) .

Rule Combination Details

The produce of one inference:

_:21 this:reqId _:19.XPath(., 92b17b0) .
_:21 this:asin _:20.XPath(tns:ASIN, 92d6890) .
_:21 this:title _:21.XPath(tns:Title, 93150e0) .

combined with another rule head:

┌ op {AWS:ItemLookup: 
│     fwrule ask(?req this:id XPath(., 91b1548) && 
│                ?req this:asin XPath(., 920be4c))
│            assert(?req this:page _:8.XPath(tns:Items/tns:Item/tns:DetailPageURL, 9127a6c) && 
│                   ?req this:reqId _:9.XPath(., 92b17b0) && 
│                   ?req this:asin _:10.XPath(tns:ASIN, 92d6890) && 
│                   ?req this:title _:11.XPath(tns:Title, 93150e0))
│               -->{AWS:ItemLookup}

produces new inferences:

_:22  this:page   _:8.XPath(this:Items/this:Item/this:DetailPageURL, 91d37c8) .
_:22  this:reqId  _:9.XPath(., 9211c38) .
_:22  this:asin   _:10.XPath(tns:ASIN, 920b2b0) .
_:22  this:title  _:11.XPath(tns:Title, 9298a50) .

Querying Rule Closure

The closure of the rules is now useful for answering user queries:

PREFIX tnss: <https://webservices.amazon.com/AWSECommerceService/2004-11-10>
SELECT ?title ?page WHERE {
_:b tnss:id "0FWYBWB91M5S26YBE382" ;
    tnss:keywords "Weaving" ;
    tnss:index "Books" ;
    tnss:asin ?asin ;
    tnss:doctitle ?title ;
    tns:page ?page }

Studying Proof Chain

The proofs of these rules illustrate how to invoke the services:

Invoking Rule Services

• Base Rules (inner-most nesting) are prooven by ground facts in the query.
  These are executed first.
• Dependent Rules use, at least in part, the product of these executions.
• These solutions produced by executing the queries replace the original solutions.
  

Documentation

• SPARQL Annotations on WSDL describes the strategy.
• AWS Choreography Example shows the working details.

Conclusions

• The SPDL annotations for AWS:ItemSearch are sufficient to execute example queries.
• The demonstrator is able to use that information to invoke operations in the correct order.
• Minimal annotation is required for automatic invocation.
• Only minimal inference required.

Future Work

There's always more to play with...

• Clean up result iteration algorithm.
• Experiment with wider range of services.
• Implement in other inference engines.
• Many potential features to add...
  

Questions/Discussion

Impressions of the demonstrator.

Applicability to industry today and tomorrow.

Applicability to users today and tomorrow.