Transcript 슬라이드 1 - Yonsei

An ontology, intelligent agent-based
framework for the provision of semantic web
services
Francisco Garcia-Sanchez, Rafael Valencia-Garcia, Rodrigo Martinez-Bejar,
Jesualdo T. Fernandez-Breis
19th May., 2009
Soft Computing Lab.
Yongjun Kim
/ 39
Outline
• Introduction
• Motivation
• The Proposed Framework
• A Proof-of-Concept Implementation
• Use Case Scenario
• Conclusion
• Future Work
• Discussion
1 / 39
Introduction
• The Intelligent Agents topic has been broadly studied and is
currently being revisited due to its relation to the semantic web.
– The semantically annotated information can be automatically
processed by agents.
• Before the semantic web, agents had to face the problems
derived from the lack of structure characterizing much of the
information published on the web.
• The Semantic Web Services are studied in order to automate
aspects of web services such as discovery, execution, selection,
composition and interoperation.
• Several research studies have shown that the cooperative
interaction between them can lead to the development of new,
more powerful applications.
2 / 39
Introduction
• The purpose of this paper:
– Highlights the problems that hamper the application of agents and
web services separately.
– Analyzes some of the projects that have attempted to combine these
technologies and point out the shortcomings of these solutions.
– Presents the SEMantic web services and Multi-Agent System
framework (SEMMAS), which provides a seamless integration of the
technologies by making use of ontologies.
• The contribution of this paper:
– An overall solution based on a fully-fledged architecture
– Proof-of-Concept implementation
• An eCommerce scenario for providing access to various computer
hardware vendors
3 / 39
Motivation
• Agent Technology
– Intelligent Agents
• An ‘agent’ is a computer system situated in some environment and
capable of autonomous action in this environment in order to meet its
design objectives (Wooldridge).
• Should have reactivity, pro-activeness, social ability (Wooldridge et al.).
• Could also have temporal continuity, reasoning, rationality, veracity,
mobility (Elamy).
– Multi-agent Systems
• Can be seen as a system consisting of a group of agents that can
potentially interact with each other.
• Presents several advantages, such as reliability, robustness, modularity,
scalability, adaptivity, concurrency, parallelism, and dynamism (Elamy).
• Can cooperate or compete.
• Can be distributed problem solving systems or open systems
(Wooldridge).
4 / 39
Motivation
• Agent Technology
– Multi-agent Systems
• Agent Communication Languages (ACL)
– KQML (Knowledge Query and Manipulation Language) and FIPA-ACL
(Foundations for Intelligent Physical Agents-Agents Communication Language)
• Agent-Oriented Software Engineering (AOSE)
– Purpose : create methodologies and tools that enable inexpensive development
and maintenance of agent-based software.
– Design Methodologies
» MAS-CommonKADS, ZEUS, GAIA, MaSE and INGENIAS
• Standardization Organizations of agent technologies
– FIPA
» Produces standards for the interoperation of heterogeneous software agents.
» Becomes an IEEE Computer Society standards organization.
» Identifies some necessary roles.
» AMS (Agent Management System), DF (Directory Facilitator), MTS (Message
Transport System).
» Has different FIPA compliant agent platform implementation.
» FIPA-OS (FIPA-Open Source), JADE (Java Agent Development Environment) and
ZEUS.
– OMG Agent PSIG
5 / 39
Motivation
• Agent Technology
– Advantages
• Appropriate for the development of complex and distributed applications.
• Can contribute features such as autonomy, proactiveness and goaloriented behavior to this kind of applications to automate some stages of
the process.
– Drawbacks
• The use of communication protocols such as IIOP (Internet Inter-Orb
Protocol) or RMI that cannot get across companies firewalls.
– Solution may create gateways between every pair of companies aimed at
interacting.
» However, no new dynamic links can be established (limited dynamism).
• The ineffective exploitation of semantics in communication mechanism.
– None of the FIPA-compliant platforms (e.g. JADE, FIPA-OS, ZEUS) provides the
proper support to handle the semantic dimension of the FIPA-ACL language.
6 / 39
Motivation
• Semantic Web Services
– The Semantic Web
• Aims at adding semantics to the data published on the Web.
– Ontologies are the back bone technology.
– Web Services
• Can be defined as services located at some point in the Internet that can
be accessed through a standard protocol.
• Based on a set of standard protocols such as UDDI (Universal
Description, Discovery, and Integration), SOAP (Simple Object Access
Protocol), and WSDL (Web Services Description Language).
• Advantages
– Enable dynamic service composition using
independent, reusable software components.
• Drawbacks
– Current technology around UDDI, WSDL,
and SOAP provide limited support
WSDL
for all that.
UDDI
search
register
SOAP
WSDL
Client
Service
Fig. 0. Web Service Architecture
7 / 39
Motivation
• Semantic Web Services
– Semantic Web Services (SWS)
• The joint application of semantic web
and web services in order to create intelligent web services.
• Standardization Activity
– OWL-S (OWL for Services)
» Makes it possible for agents to discover, compose, invoke, and monitor services
with a high degree of automation.
» Composed of three main parts:
» The service profile for advertising and discovering services
» The process model which gives a detailed description of a service operation
» The grounding which provides details on how to interoperate with a service via
messages
– WSMO (Web Service Modeling Language)
» Provides a conceptual framework for semantically describing all relevant aspects
of web services in order to facilitate the automation of discovering, combining
and invoking electronic services over the Web.
» Comprised of three different elements:
» Web Service Modeling Ontology (WSMO)
» Web Service Modeling Language (WSML)
» Web Service Execution Environment (WSMX).
8 / 39
Motivation
• Semantic Web Services
– Semantic Web Services (SWS)
• Standardization Activity
– SWSF (SWS Framework)
» Composed of the SWS Language (SWSL) and the SWS Ontology (SWSO).
» Two formal characterization of the model have been identified:
» FLOWS (First-Order Logic Ontology for Web Services)
» SWSL-FOL
» ROWS (Rules Ontology for Web Services)
– WSDL-S (WSDL Semantics)
» Defines a mechanism to associate semantic annotations with Web Services that
are described using WSDL.
» The semantic information includes definitions of precondition, input, output, and
effects on web service operations.
– SAWSDL (Semantic Annotations for WSDL)
» Identifies some WSDL and XML Schema extension attributes that support the
semantic description of WSDL components.
» Necessary to link WSDL elements with semantic models, such as ontologies.
9 / 39
Motivation
• Semantic Web Services
– Semantic Web Services (SWS)
• SWS-related Tools
– WSMX (Web Services Execution Environment)\
» An execution environment to perform dynamic discovery, selection, mediation,
invocation and interoperation of SWS.
– IRS (Internet Reasoning Service)
» Allows applications to semantically describe and execute web services.
» Supports the provision of semantic reasoning services within the context of the
semantic web.
– SWS Limitations
• SWS technologies depend on higher-level software entities with cognitive
capabilities able to access the semantic content of service descriptions,
process it and understand it.
10 / 39
Motivation
• Intelligent Agents and SWS: related work
– Numerous research projects have been carried out due to:
• The shortcomings of intelligent agent technologies.
• The inherent need for autonomous software entities in SWS
environments.
• The promising benefits of having two technologies working cooperatively.
– Three main scenarios are possible:
• Agents over Web Services
• Agents in Web Service Wrappers
– Communication in web services and agents becomes equivalent, so the
distinction between them disappears.
• Agents and Web Services
– Both types remain separate interoperating through gateways and translation
processes.
11 / 39
Motivation
• Intelligent Agents and SWS: related work
– [Hendler, 2001]
• Description
– Describes at first how the ontology languages of the semantic web could lead to
more powerful agent-based approaches to using services offered on the Web.
– Proposes a method for describing the way the invocation of services should be
done by agents by means of an ontology language such as DAML+OIL.
– The Semantic Web FRED project (SWF) [Stollberg et al., 2005]
• Description
– Combines agent technology, ontologies, and SWS in order to develop a system
for automated cooperation.
– Is a complete solution for the interaction agent-service and provides the means
for performing tasks such as service discovery, composition and execution.
– The authors distinguish three types of services: plans (Java programs),
processes (complex and nested services), and external web services (through
WSDL).
• Drawbacks
– It appears to be tightly bound to WSMO (a kind of WSMO implementation).
– Lacking a common accepted standard for SWS, the attachment of SWF to a
particular approach seems too risky.
12 / 39
Motivation
• Intelligent Agents and SWS: related work
– GODO (Goal-Oriented Discovery for SWS) [Gomez et al., 2006]
• Description
– Uses the ontology to generate the goals to be executed and sends them to the
appropriate execution environment.
» Incorporates a language analyzer that determines the concepts, attributes,
attribute values, and relationships within a sentence.
• Drawbacks
– Interacts with SWS infrastructures instead of the SWS themselves. It can’t benefit
from most of the advantages agent technology provides in these environments
» Agent-based negotiation can’t take place.
» Agent techniques have to be implemented ad hoc within the infrastructure.
– [Bhuler and Vidal, 2005]
• Description
– Uses an intelligent agent acting on behalf of a web service within a workflow.
» When a web service is intended to be executed, its representative agent gets the
control flow and attempts to improve the workflow process.
– Envisions workflow-based MAS where service providers are agents themselves,
thus acquiring the full proactive, autonomous, and selfish characteristics.
• Drawbacks
– Semantically described web services are not considered at all.
13 / 39
Motivation
• Intelligent Agents and SWS: related work
– Agents and Web Services Interoperability Working Group (AWSI WG)
• Description
– Part of the IEEE FIPA Standards Committee.
– Creates a middleware able to handle the fundamental differences between agent
technology and web services.
» The use of different communication protocols (ACL vs. SOAP)
» Service description languages (DF-Agent-Description vs. WSDL)
» Service registration mechanisms (DF vs. UDDI)
– AgentWeb Gateway Middleware [Shafiq et al., 2006]
• Description
– Facilitates the required integration without changing existing specifications and
implementations of both technologies.
– Agents and web services remain at the same abstraction level creating a
heterogeneous service space and interoperating through the gateway, thus
complying with the third category of Blacoe and Protabella’s classification.
» Focuses on the overlapping features of the technologies.
• Drawbacks
– Each of these technologies must be situated at a different abstraction level
since most of the functionalities provided by Intelligent Agents and Web
Services is complementary.
14 / 39
The Proposed Framework
• Previous approaches suffer from shortcomings mainly due to
their inability to overcome the problems associated to each of the
technologies under question.
• Attempts to solve some of these problems by means of the
framework, which is based on a loosely coupled infrastructure
and makes use of ontologies to facilitate agents and services
interoperation.
• The conceptual differences between agent technology and
(semantic) web services technologies lead to the need to have
both technologies working in an integrated environment.
– Agent technologies aim to act as autonomous entities.
– Web services aim to provide worldwide-accessible functionality.
15 / 39
The Proposed Framework
• Foundations
– Ontologies operate as the ‘glue’ that binds other components.
• Function as universal vocabularies so that web services and agents share
the same interpretation of the terms contained in the messages.
• Useful to describe web service capabilities and processes.
• Extract or build the local domain-related knowledge of agents.
• Work as negotiation protocols and strategies between agents.
16 / 39
The Proposed Framework
• Foundations
– Based on a multi-tier architecture that is composed of four different
layers.
• Business logic layer, semantic web services layer, intelligent agents layer
and application layer.
New services can emerge,
A service may change their functionalities
or can even disappear at run-time.
17 / 39
The Proposed Framework
• The SEMMAS (SEMantic web services and MAS) Framework
– Becomes independent from both the domain and the actual
application since it comprises only the intelligent agents layer and
the semantic web services layer.
– The architecture is composed of three components:
• A set of intelligent agents, four ontology repositories, and three interfaces
18 / 39
The Proposed Framework
• The SEMMAS (SEMantic web services and MAS) Framework
– Agents
• Seven types of agents are grouped in three main categories:
– Agents that act on behalf of service owners
» Provider agent and service agent
» Manage the access to services and ensure that the contracts are fulfilled.
– Agents that act on behalf of service consumers
» Customer agent, discovery agent, and selection agent
» Locate services, agree on contracts, and receive and present results.
– Agents that perform management tasks
» Framework agent and broker agent
» Avoid system resources becoming overloaded.
» Monitor the status of all the interactions.
– Roles
• Are encapsulations of dynamic behaviors and properties that can be
played by agents.
• Are dynamic and flexible, responsibility-driven, and context-sensitive.
• Two categories of roles are identified.
– Roles dealing with service-related issues
– Roles related to the framework management
19 / 39
The Proposed Framework
• The SEMMAS (SEMantic web services and MAS) Framework
20 / 39
The Proposed Framework
• The SEMMAS (SEMantic web services and MAS) Framework
21 / 39
The Proposed Framework
• The SEMMAS (SEMantic web services and MAS) Framework
22 / 39
The Proposed Framework
• The SEMMAS (SEMantic web services and MAS) Framework
– Ontology Repositories
• Four kinds of ontologies have been included within the platform.
• Application and Domain Ontology
– Supports the communication among the components in the framework.
» Application ontology contains the knowledge entities (i.e. concepts, attributes,
relationships, and axioms) that model the application.
» Domain ontology represents conceptualization of the specific domain.
• Agent Local Knowledge Ontology
– Contains the knowledge about the environment an agent possesses.
– Includes knowledge about the assigned tasks, as well as mechanisms and
resources available to achieve those tasks.
» E.g., Broker Agent’s one may contain the mapping rules it has to apply to resolve
the interoperability mismatches.
• Negotiation Ontology
– Comprises both negotiation protocols and strategies that constitute the
negotiation mechanisms agents must use to coordinate their interactions.
• Semantic Web Services Ontologies
– Contain the semantic description of Web Services.
23 / 39
The Proposed Framework
• The SEMMAS (SEMantic web services and MAS) Framework
– User Interfaces
• Is a key component of the infrastructure.
• Three different interfaces are included:
– Customization Interface (for software developers)
» Add/delete ontology repositories (need URLs), add/delete ontologies, roles
implementation, agent instantiation
– Service Providers Interface
» Add/delete services, set general preferences (conditions), set service particular
preferences
– Service Consumers Interface
» Set general preferences (e.g. expected cost, warranty, desired delivery time),
query, execute services
24 / 39
A Proof-of-Concept Implementation
• A MAS was implemented on top of the JADE platform.
• Methodology
– INGENIAS was applied to design the multi-agent infrastructure.
– Proposes the MAS analysis and design from five points of view:
• Organization :
– Defines how the agents are grouped, the system functionality, and what
restrictions should be imposed over the agent behavior.
• Agent
– Describes the agents themselves and the mental states they have along their life
cycle.
• Tasks and Objectives
– Supports to analyze and define the system objectives, their decomposition, and
their relation with tasks.
• Interactions
– Shows the details about how the agents coordinate and communicate.
• Environment
– Defines what is around the system and how each agent perceives it.
25 / 39
A Proof-of-Concept Implementation
• INGENIAS Methodology
– Composed of three phases.
• Inception, elaboration, and constructions
– Supports the INGENIAS Development Kit (IDK).
• It provides the means to create most of the diagrams and models
required by the methodology.
• It incorporates a function to automatically generate JADE files from the
developed diagrams.
– Advantages of JADE
» Complies with the FIPA specification, open-source, implemented in Java, and at
a mature state (version 3.4).
26 / 39
A Proof-of-Concept Implementation
• INGENIAS Methodology
27 / 39
A Proof-of-Concept Implementation
• INGENIAS Methodology
28 / 39
Use Case Scenario
• eCommerce domain is used to evaluate the prototype.
– Provides support for B2B and B2C transactions in an environment
where various computer hardware vendors are present.
– Needs to select the best proposal according to the user’s
preferences.
• Negotiation is a key issue in these environments.
• eCommerce Scenario
– Comprises of three computer hardware providers whose products
catalogues are accessible through three different Web Services.
• Provider A-C, Catalog A-C, Web Service A-C
– Service providers have established a set of preferences.
• Preferences A-C
– Each provider makes use of a different ontology.
• Ontology A-C
– The prototype uses an ontology.
• Ontology S
29 / 39
Use Case Scenario
• eCommerce Scenario
30 / 39
Use Case Scenario
• Prototype Customization
– Four main tasks for customizing the framework to a particular
application domain
• Designs a domain ontology for PC-components in OWL containing the
concepts and relationships.
• Develops a web service for each provider.
• Annotates the services semantically by using the providers ontologies.
• Instantiates the agents that are necessary to provide access to computer
hardware catalogues through the available services.
– Domain Ontology
• Has been written in OWL-DL by using Protégé.
• Consists of 45 concepts, 8 attributes, and 55 relationships (43
taxonomical and 12 meteorological).
– Taxonomical : ‘classes’, ‘datatype properties’, ‘hierarchical sub-classes’
– Meteorological : ‘part of’
31 / 39
Use Case Scenario
• Prototype Customization
– Domain Ontology
32 / 39
Use Case Scenario
• Prototype Customization
– Semantic Web Services
• Simulate the internal business processes of the companies meeting the
needs around B2B and B2C transactions.
• Java and the Apache Axis2 library have been used to implement services.
• OWL-S approach is used to describe the service.
– Utilize OWL-S Editor, a Protégé plugin.
– Semantic web service C
» ‘Provider C’ owns the ‘web service C’, which is composed of a method
‘buyProduct_C’.
» buyProduct_C(Product att, int no) → Product
• WSDL description is needed to allow the service be accessed.
– However, it supports a mere syntactic description of the operations.
• A formal semantic description is needed to allow the service be
automatically exploited.
– OWL-S is applied here.
• The annotation of the services allows the software agents to interact with
providers on behalf of human users.
33 / 39
Use Case Scenario
• Prototype Customization
– Semantic Web Services
34 / 39
Use Case Scenario
• Prototype Customization
– A user wants to acquire a 4GB USB flash drive.
• Providers and services identification
– Processes and interprets the user request.
» Obtains an internal, formal representation of the user goal.
– Customer Agent sends it to Discovery Agent.
– Discovery Agent returns the list to Customer Agent, which forwards it to
Selection Agent.
• Negotiation and contract preparation
– Selection Agent gets in contact with Service Agent.
– Service Agent requests the corresponding Provider Agent for the high-level,
strategic preferences, and sends back to Selection Agent.
» Payment types, delivery time and warranty period
– Selection Agent compares each service preferences with the user preferences
and calculates their expected utility.
• Negotiation and contract preparation
– Selection Agent sorts the list of matched services and sends it back to
Customer Agent.
35 / 39
Use Case Scenario
• Prototype Customization
– A user wants to acquire a 4GB USB flash drive.
• Transaction fulfillment
– The system provides the means to successfully perform the transaction, that is,
execute the chosen service.
– Customer Agent receives the selection, and starts interacting with the
corresponding Service Agent.
– Service Agent receives and analyzes the service semantic description.
– The service is invoked and the outcome is sent back to Customer Agent.
• Product/service evaluation and contract fulfillment
– Framework Agent carries out a high-level monitoring process, while Customer
and Service Agents ensure their counterparts fulfill their sides of the agreement.
36 / 39
Conclusion
• The advantages and drawbacks of agent technology and web
services were discussed.
• Joining together intelligent agents and semantic web service
technologies seemed to be able to lead to the development of
new, more powerful applications.
• SEMMAS, an ontology-based framework for intelligent agents
and semantic web service integration was introduced.
– Ontologies constitute the facilitating technology that enables a
seamless communication between agents and services.
• The purpose of this work was far however from procuring new
innovative solutions for semantic web services related tasks.
– Matchmaking, composition, mediation, and so on.
– Bioinfomatics, eCommerce, eGovernment (focus of our research)
37 / 39
Future Work
• Evaluate the framework in terms of its performance and usability
in several domains.
• A number of algorithms for web service discovery, composition
and invocation will be tested and integrated into the
implementation.
• The performance and accuracy of various ontology reasoners
and inference engines will be examined and the most effective
ones included into the platform.
• Perform the eventual integration of grid services within the
framework.
38 / 39
Discussion
• Can Web Services solve the problem related to the company
firewalls?
– Even UDDI may have problems in accessing services behind
firewalls.
• Does it make sense to prefer the adoption of Web Services to the
modification of any FIPA-compliant platform for handling the
semantic dimensions of FIPA-ACL language?
39 / 39