Agent Model for Interaction with Semantic Web Services

Ivo Mihailovic

Overview

Introduction About Semantic Web About Web Services About Semantic Web Services The model of the agent eGovernment example

Introduction

An attempt to demystify the Semantic Web to the author

Agent for interaction with Semantic Web Services

Integration of eGov services: providing a single access point to all government services, the increase in their interoperability without affecting their autonomy, approach oriented towards citizens

Semantic Web

Web Services

The W3C defines a Web service as a software system designed to support interoperable Machine to Machine interaction over a network.

Require three basic components: A messaging service for communicating between resources on the Web (SOAP), An interface description language (WSDL), A registration service so that clients looking for a service can find them (UDDI).

Semantic Web Services

Semantic Web Services are self-describing, semantically marked-up software resources that can be interacted with in a task driven automatic way.

Dynamic part of the semantic web.

Semantic Web Services

Automated Web Service Discovery Automated Web Service Invocation Automated Web Service Composition Automated Web Service Monitoring Automated Web Service Verification Automated Web Service Simulation

Semantic Web Services

WSDL can specify the operations available through a web service and the structure of data

WSDL cannot specify semantic meaning of the data or semantic constraints on the data.

WSMO vs OWL-S

WSMO

WSMO

Ontologies – describe all relevant concepts and relations among them

Web Services - describe capabilities, interfaces and internal working of the Web service

Goals - describe the capability the user would like to have and the interface he would like to interact with

Mediators - define mappings between components

OWL-S

OWL-S Service Profile

Service Profile – describes what the service does, used for the discovery

Intended to be published in a WS registry Classification through the creation of a subclass

hierarchy (one WS can belong to multiple classes) IOPEs represent Web service’s capabilities, they

describe the service Other features: contact information, category, quality

rating etc.

OWL-S Process Model

Process Model – describes how the service works, its operation, control and data flow

Used for selection, invocation, interoperation, composition, and monitoring of the WS

Process types: atomic, simple, composite

OWL-S Grounding

Grounding specifies how to access the service in terms of communication protocols and message descriptions

Mapping from Process Model to WSDL

Agent model

Example: The user wants to book a flight. He should find a service that sells tickets, check if it

accepts his credit card and book the flight The agent can automate this using an ontology of

Web services Agent tries to resolve the semantics of the user with

the semantics of the service User provide goals, agent try to map them to actions

Agent model

Example: Goal – to go on a business trip. Actions – to book a flight, a taxi to the hotel and a hotel room

Example: Goal – the change of address. Agent needs two ontologies: First to map user’s requests to available

goals Second to map goals to Web services

Scenario of operation

Step 1) User enters the query

Step 2) Agent maps user’s requirements to goals

The hard part – needs NLP, needs to be able to interpret user’s constraints

Book me a flight if the weather’s nice.

Scenario of operation

Step 3) Agent tries to discover a service for each goal

Step 4) If it cannot find an integrated service for a goal it tries to decompose it

Step 5) After making the plan of actions the agent invokes the services

Area of application: e-Gov

Integration of services, increased interoperability, not affection the autonomy of government organizations

Government is a dynamic area, things constantly changing (especially in countries in transition)

Services easily changed, added or removed

Area of application: e-Gov

E-Government Domain Ontology encodes organizational, legal, economic, business etc. concepts

Web services ontology should categorize services to enable discovery

Registry of Web services with the service descriptions

A change of the service requires the update of the description

Service discovery

WSMO – comparing Goals to Web Service descriptions

OWL-S – using the Service Profile Service classification by creating a hierarchy

of subclasses of Service Profile SWS yellow pages – a class-hierarchical

taxonomy

Service discovery

Classification by serviceProduct and serviceCategory properties

Mapping to an OWL specification of UNSPSC (United Nations Standard Products and Services Code)

Also can be connected to a classification outside of OWL

Service composition

AI planning to form service composition A planning problem P is a 3-tuple < I, G, A > I – description of the initial state G – description of the goal state A – set of actions for state transformations An action sequence (a plan) S is a solution to

P if S can be executed from I and the resulting state of the world contains G.

Service composition

States formed from user’s goals and services preconditions and results

Actions are descriptions of Web services which show how their execution modifies the state of the world

STRIPS algorithm

E-Gov ontology

Needs to represent the viewpoint of citizens as well as government

Connecting terms from common language with their legal jargon counterparts

Descriptions of non web-based services as well

Planner could make hybrid composition

Thank you!