sewasie: a semantic search engine sonia bergamaschi, maurizio vincini università di modena e reggio...

SEWASIE: a Semantic Search Engine

Sonia Bergamaschi, Maurizio VinciniSonia Bergamaschi, Maurizio Vincini

UniversitUniversità di Modena e Reggio Emiliaà di Modena e Reggio Emilia

21-22 October 2002 Vilnius, LithuaniaVilnius, Lithuania

TELEBALT Conference Teleworking for Business, Education, Research and e-CommerceTeleworking for Business, Education, Research and e-Commerce

Sonia Bergamaschi – Università di Modena e Reggio Emilia 2

SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

OutlineOutline

What is SEWASIE?What is SEWASIE?– ObjectivesObjectives– Expected resultsExpected results– Main InnovationsMain Innovations– The high level ArchitectureThe high level Architecture– ComponentsComponents– A SINodeA SINode

A P2P paradigm for SEWASIE (for later discussion)A P2P paradigm for SEWASIE (for later discussion)– INTER SINode NetworkINTER SINode Network– Brokering Agent NetworkBrokering Agent Network


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

SEWASIESEWASIE SEWASIE (Semantic Webs and AgentS in Integrated Economies) is a SEWASIE (Semantic Webs and AgentS in Integrated Economies) is a

research project founded by EU on action line Semantic Web (May research project founded by EU on action line Semantic Web (May 2002/April 2005)2002/April 2005)

http://www.sewasie.orghttp://www.sewasie.org The consortium detailsThe consortium details

Università degli Studi di Modena e Reggio EmiliaUniversità degli Studi di Modena e Reggio Emilia (ITALY) (ITALY) CNA SERVIZI Modena s.c.a.r.l.CNA SERVIZI Modena s.c.a.r.l. (ITALY) (ITALY) Università degli Studi di Roma “La Sapienza”Università degli Studi di Roma “La Sapienza” (ITALY) (ITALY) Rheinisch Westfaelische Technische Hochschule AachenRheinisch Westfaelische Technische Hochschule Aachen ( (GERMANYGERMANY)) Libera UniversitLibera Università di Bolzano (ITALY)à di Bolzano (ITALY) Thinking Networks AGThinking Networks AG ( (GERMANYGERMANY)) IBM Italia SPAIBM Italia SPA (ITALY) (ITALY) Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung eingetragener VereinFraunhofer-Gesellschaft zur Förderung der angewandten Forschung eingetragener Verein

((GERMANYGERMANY))

Contact information:Contact information:

Prof. Sonia BergamaschiProf. Sonia Bergamaschi

DII – UniversitDII – Universitàà di Modena e Reggio Emiliadi Modena e Reggio Emilia

Tel: +39 059 2056132 Fax: +39 059 2066126Tel: +39 059 2056132 Fax: +39 059 2066126

bergamaschibergamaschi..soniasonia@@unimounimo.it.it

http://www.dbgroup.unimo.it/Bergamaschi.htmlhttp://www.dbgroup.unimo.it/Bergamaschi.html


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

SEWASIE ObjectivesSEWASIE Objectives

Design and implementation of an advanced search engineDesign and implementation of an advanced search engine

The SEWASIE project pursues the following aims: To develop an agent-based secure, scalable and distributed system architecture for semantic search (based on ontologies) and for structured web-based communication. To provide semantic enrichment processes for knowledge-based extraction of meta-information of heterogeneous data sources. To develop a general framework for query management and information reconciliation based on a semantically enriched data and trusted agent structure. To develop an information brokering component which includes methods for collecting, contextualising and visualising semantically rich data. To develop communication processes that enable the use of multilingual ontologies. To provide the end-user with efficient interfaces for formulating queries using a graphical representation and for intelligent navigation through the semantically enriched information space.


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

Expected ResultsExpected Results

• In particular, In particular, SEWASIE has toSEWASIE has to:: Help European SMEs to find the right strategic information at the Help European SMEs to find the right strategic information at the

right time in a multinational environment;right time in a multinational environment; Provide advanced and novel services for monitoring and linking Provide advanced and novel services for monitoring and linking

information in the context of risk management and competitor information in the context of risk management and competitor analysis;analysis;

Provide ontology-based communication mechanisms for Provide ontology-based communication mechanisms for negotiation in multi-language environments;negotiation in multi-language environments;

Ease the use of complex cross-language retrieval and data Ease the use of complex cross-language retrieval and data condensation tools by providing intuitive interfaces.condensation tools by providing intuitive interfaces.

The SEWASIE vision helps European enterprises to compete in a global market and to form strategic alliances at a European level by providing a sophisticated retrieval, brokering and communication service on basis of the semantic web technology.


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

The very high level architectureThe very high level architecture

User Interface

Query Agent

SEWASIEInformation

Node

VirtualDataStore

SEWASIEInformation

Node

VirtualDataStore

Query Agent Query Agent

SEWASIEInformation

Node

VirtualDataStore

SEWASIEInformation

Node

VirtualDataStore

SEWASIEInformation

Node

VirtualDataStore

Brokering Agent Brokering Agent


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

The very high level architectureThe very high level architecture

Tools and methods has to be developed to create/maintain Tools and methods has to be developed to create/maintain multilingual ontologies, with an inference layer grounded in W3C multilingual ontologies, with an inference layer grounded in W3C standards (XML, XML Schema, RDF(S)).standards (XML, XML Schema, RDF(S)).

Search results will be personalised and visualised according to Search results will be personalised and visualised according to users’ preferences. users’ preferences.

From an architectural point of view, SEWASIE aims to provide an From an architectural point of view, SEWASIE aims to provide an open and distributed architecture based on intelligent agents open and distributed architecture based on intelligent agents (brokers, mediators and wrappers) facing scalability and (brokers, mediators and wrappers) facing scalability and flexibility issues, i.e. the ability to fit in changing and growing flexibility issues, i.e. the ability to fit in changing and growing environments and to interoperate with other systems, while environments and to interoperate with other systems, while offering one central point of access to the user.offering one central point of access to the user.

The main actors on stage areThe main actors on stage are The user interfaceThe user interface The query agentThe query agent The brokering agentThe brokering agent The information node (SINode)The information node (SINode)


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

Main Innovations (1)Main Innovations (1)

The SEWASIESEWASIE project aims to develop an advanced search engine enabling intelligent access to heterogeneous data sources on the web, via semantic enrichment, to provide the basis for structured web-based communication.

The SEWASIE system will realise a virtual network, SEWASIE Virtual Network (SVN) whose nodes are SEWASIE Information Nodes (SINode).

– SINodes are multi-database mediator-based systems, each including a Virtual Data Store, an Ontology Builder, and a Query Manager

– The managed Information Sources are heterogeneous collections of structured, semi-structured, or unstructured data, e.g. relational databases, XML or HTML documents

multilingual ontologies and agents


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

Main Innovations (2)Main Innovations (2)

– Ontologies are multilingualOntologies are multilingual

– The Brokering Agent/Agents maintains the knowledge related The Brokering Agent/Agents maintains the knowledge related to the SEWASIE Virtual Network and the user profiles. to the SEWASIE Virtual Network and the user profiles.

– In the SEWASIE Virtual Network, the Brokering Agent classifies In the SEWASIE Virtual Network, the Brokering Agent classifies SINodes, it is responsible for handling the acquisition of a new SINodes, it is responsible for handling the acquisition of a new SINode and for consequently updating of the SEWASIE Virtual SINode and for consequently updating of the SEWASIE Virtual Network.Network.

– In query solving phase, starting from a specified SINode, the In query solving phase, starting from a specified SINode, the Query Agent accesses other SINodes and thus collects partial Query Agent accesses other SINodes and thus collects partial answers.answers.

– To select SINodes useful to solve a query, a Query Agent To select SINodes useful to solve a query, a Query Agent interacts with a Brokering Agent. interacts with a Brokering Agent.


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

Sewasie ArchitectureSewasie Architecture


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

Components (actors and stages)Components (actors and stages)

The basic (generic) user query scenario we have in mind concerns a user at a The basic (generic) user query scenario we have in mind concerns a user at a workstation (or handling a handheld computer, or a cellular phone with network workstation (or handling a handheld computer, or a cellular phone with network connection capabilities), looking for information on a topic. The user may then connection capabilities), looking for information on a topic. The user may then issue a request expressed in some “natural” language style to the network. The issue a request expressed in some “natural” language style to the network. The user interface translates the user request into a query, keeping into account the user interface translates the user request into a query, keeping into account the past history and present context of the user, and sends a probe out (the query past history and present context of the user, and sends a probe out (the query agent, QA) scouting for answers. agent, QA) scouting for answers.

The QA connects into the network of SEWASIE brokering agents (BAs) and queries The QA connects into the network of SEWASIE brokering agents (BAs) and queries them for info on the matter of interest. A typical interaction between a QA and a them for info on the matter of interest. A typical interaction between a QA and a BA may imply that the BA will provide directions to relevant SINodes and BA may imply that the BA will provide directions to relevant SINodes and information on SINode contents, or reference the QA to other BAs. The QA will then information on SINode contents, or reference the QA to other BAs. The QA will then move to such nodes and query them, or may move on to the other BAs to ask move to such nodes and query them, or may move on to the other BAs to ask them for directions again. them for directions again.

When the QA receives the SINode answers, it has to integrate them, possibly When the QA receives the SINode answers, it has to integrate them, possibly querying some BA again (data reconciliation).querying some BA again (data reconciliation).

Another type of user query is longer-term network monitoring request. While the Another type of user query is longer-term network monitoring request. While the previous one is a short-term straight request which terminates with the return of previous one is a short-term straight request which terminates with the return of answers or the decision that there aren’t any, the monitoring request is an open-answers or the decision that there aren’t any, the monitoring request is an open-ended request for information available which rather looks for changes in the ended request for information available which rather looks for changes in the content of the network. In this case the QA will monitor a certain predefined view content of the network. In this case the QA will monitor a certain predefined view of the domain. The QA will return to the user interface any change that will be of the domain. The QA will return to the user interface any change that will be detected over time.detected over time.


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

The SINode moduleThe SINode module SINodesSINodes are mediator-based are mediator-based systems, including: systems, including: A A Virtual Data StoreVirtual Data Store (VDS) (VDS) represents a virtual view of the represents a virtual view of the overall information managed overall information managed within any SINode and consists within any SINode and consists of the managed information of the managed information sources, wrappers, and a sources, wrappers, and a metadata repository.metadata repository. The managed The managed Information Information SourcesSources are heterogeneous are heterogeneous collections of structured, semi-collections of structured, semi-structured, or unstructured structured, or unstructured datadata. .

…

Ontology

StructuredDatabases

RDBs

Wrapper

SemanticEnrichment

QueryManager

MetadataRepository

Semi-structuredDatabases

Wrapper

SemanticEnrichment

<XML><DATA>...</DATA>

Wrapper

SemanticEnrichment

UnstructuredText documents

<HTML>...

Ontology Builder

Virtual Data Store

Ontology

StructuredDatabases

RDBs

Wrapper

SemanticEnrichment

QueryManager

MetadataRepository


Wrapper

SemanticEnrichment


Wrapper

SemanticEnrichment


<HTML>...

Ontology Builder

Virtual Data Store

A A WrapperWrapper implements common communication protocols and translates implements common communication protocols and translates to and from local access languages. There is one wrapper linked to each to and from local access languages. There is one wrapper linked to each information source. information source. The The Ontology BuilderOntology Builder performs semantic enrichment processes in order performs semantic enrichment processes in order to create and maintain the currentto create and maintain the current Ontology Ontology which is made up of the which is made up of the Global Virtual View of the sources and the mapping description between the Global Virtual View of the sources and the mapping description between the GVV itself and the sources.GVV itself and the sources. The The Metadata Repository Metadata Repository holds the ontology and the knowledge holds the ontology and the knowledge required to establish semantic relationships between the SINode itself and required to establish semantic relationships between the SINode itself and the neighbouring ones.the neighbouring ones.


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

Virtual Data StoreVirtual Data Store

Global VDS model and language The first tenet of the architecture within the VDS is a common model and the associated

languages, travelling as payload on the global VDS infrastructure. The main requirements for a candidate language are

a rich syntax for ontology description, including mapping relations GVV/Sources a flexible query language and tools for effective translation of queries and results

among modules. One candidate for the data model and associated languages is ODMI3 (ODLI3), which was

derived from ODMG specification; one candidate for the query language is OQLI3. Notice that the adoption of specific languages for intra-node communication does not avoid

to put at the SEWASIE network disposal the information managed by SINode in other format.

…

Ontology

StructuredDatabases

RDBs

Wrapper

SemanticEnrichment

QueryManager

MetadataRepository


Wrapper

SemanticEnrichment


Wrapper

SemanticEnrichment


<HTML>...

Ontology Builder

Virtual Data Store

Ontology

StructuredDatabases

RDBs

Wrapper

SemanticEnrichment

QueryManager

MetadataRepository


Wrapper

SemanticEnrichment


Wrapper

SemanticEnrichment


<HTML>...

Ontology Builder

Virtual Data Store


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

Virtual Data StoreVirtual Data Store

Global VDS infrastructure The architecture of this module is inherently distributed (i.e. in most cases its

functionality will be distributed among several host machines of different types). As a consequence, these components will all need to have inter-process communication functionalities to support the interaction. The first choice here is to use the TCP/IP family of protocols, which are universally supported at all levels. Above the basic network layer we need to select a proper enveloping mechanism to guarantee the higher level properties of the communication.

We want to have– verified point-to-point communications– no special requirements to pass across common boundaries like firewalls (at least

those with typical policy definitions)– an option to use reserved (encrypted) communications– standardisation and widespread availability (at least in a medium term perspective)

A natural candidate for such a protocol family is provided by the SOAP/WSDL environment. An alternative to the SOAP/WSDL/UDDI family is given by the CORBA architecture. Based on these protocols the VDS will have an API made available to applications (agents or others) in order to

– Query the content of the SINode: receive a query and return the corresponding results– Manage the semantic profile of the SINode in the SEWASIE network: keeping the up to

date semantic profile, and updating it


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

WrappersWrappers

Wrappers are the “docking stations” of the heterogeneous data sources contributing content to SEWASIE. They are software modules in charge of the mediation between the internals of each data source and the functionalities of the SINode.

Different wrappers have to be defined to cover structurally diverse sources. The internals of the wrapper will need to be modular. However, the interface of these modules

will be uniform and independent of the underlying source type. Two major functions need to be performed by these wrappers:

– to support the translation of the structure of the information managed by local sources into the SINode description language ODLI3

– to support the translation of the queries from the SINode query language OQLI3 into the specific query language of the underlying source.

To this aim, functionalities and protocols will need to be made available in order to enable the communication between wrappers and the Query Manager and the Ontology Builder.

…

Ontology

StructuredDatabases

RDBs

Wrapper

SemanticEnrichment

QueryManager

MetadataRepository


Wrapper

SemanticEnrichment


Wrapper

SemanticEnrichment


<HTML>...

Ontology Builder

Virtual Data Store

Ontology

StructuredDatabases

RDBs

Wrapper

SemanticEnrichment

QueryManager

MetadataRepository


Wrapper

SemanticEnrichment


Wrapper

SemanticEnrichment


<HTML>...

Ontology Builder

Virtual Data Store


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

Ontology BuilderOntology Builder

The Ontology Builder (OB) is the collective name of a set of functionalities which will support the creation and maintenance of the GVV of the SINode. Given common model and languages, we need to establish tools for synthesizing ontologies and merging them into a GVV, with the final goal of developing a shareable ontology at the SINode level.

The ontology building process is a cooperative one, involving the designers, the wrappers of the sources providing raw data to the OB, which performs the integration, saves the results in the Metadata Repository, and publishes them to the BAs.

The building process begins with the creation of a common thesaurus of the information provided by wrappers, that is terminological intensional and extensional relationships describing intra-schema knowledge about classes and attributes of each source schemas.

…

Ontology

StructuredDatabases

RDBs

Wrapper

SemanticEnrichment

QueryManager

MetadataRepository


Wrapper

SemanticEnrichment


Wrapper

SemanticEnrichment


<HTML>...

Ontology Builder

Virtual Data Store

Ontology

StructuredDatabases

RDBs

Wrapper

SemanticEnrichment

QueryManager

MetadataRepository


Wrapper

SemanticEnrichment


Wrapper

SemanticEnrichment


<HTML>...

Ontology Builder

Virtual Data Store


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

Query ManagerQuery Manager

The Query Manager is the coordinated set of functions which take an incoming query, define a decomposition of the query according with the mapping of the global virtual view of the SINode onto the specific data sources available (GAV approach) and relevant for the query, sends the queries by means of local QAs to the wrappers in charge of the data sources, collects their answers, performs any residual filtering as necessary, and finally delivers whatever is left to the requesting query agent.

…

Ontology

StructuredDatabases

RDBs

Wrapper

SemanticEnrichment

QueryManager

MetadataRepository


Wrapper

SemanticEnrichment


Wrapper

SemanticEnrichment


<HTML>...

Ontology Builder

Virtual Data Store

Ontology

StructuredDatabases

RDBs

Wrapper

SemanticEnrichment

QueryManager

MetadataRepository


Wrapper

SemanticEnrichment


Wrapper

SemanticEnrichment


<HTML>...

Ontology Builder

Virtual Data Store


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

Ontology BuilderOntology Builder

Based on such information and on designer supplied relationships capturing specific domain knowledge, the OB performs semiautomatic inter-schema analysis by: – exploiting lexicon derived relationships, which are based on processes like

synonyms identification or generalisation-specialisation relations, and– inferring new relationships.

All these relationships are considered in the subsequent phase of construction of the ontology. Such an activity is based on hierarchical clustering techniques and supports the emergence of a number of global classes (GVV) representative of all the classes coming from the sources and of a mapping description between the GVV and the local sources.

Most of ideas comes from the MOMIS project http://www.dbgroup.unimo.it/Momis


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

The MOMIS project (bibliografy)The MOMIS project (bibliografy)

S. Bergamaschi, S. Castano e M. Vincini "Semantic Integration of S. Bergamaschi, S. Castano e M. Vincini "Semantic Integration of Semistructured and Structured Data Sources", SIGMOD Record Special Semistructured and Structured Data Sources", SIGMOD Record Special Issue on Semantic Interoperability in Global Information, Vol. 28, No. 1, Issue on Semantic Interoperability in Global Information, Vol. 28, No. 1, March 1999March 1999

D. Beneventano, S. Bergamaschi, S. Castano, A. Corni, R. Guidetti, G. D. Beneventano, S. Bergamaschi, S. Castano, A. Corni, R. Guidetti, G. Malvezzi, M. Melchiori e M. Vincini: "Information Integration: the MOMIS Malvezzi, M. Melchiori e M. Vincini: "Information Integration: the MOMIS Project Demonstration", International Conference on Very Large Data Project Demonstration", International Conference on Very Large Data Bases (VLDB'2000), Cairo, Egypt, Settembre 2000Bases (VLDB'2000), Cairo, Egypt, Settembre 2000

S. Bergamaschi, S. Castano, D. Beneventano e M. Vincini: "Semantic S. Bergamaschi, S. Castano, D. Beneventano e M. Vincini: "Semantic Integration of Heterogeneous Information Sources", Special Issue on Integration of Heterogeneous Information Sources", Special Issue on Intelligent Information Integration, Data & Knowledge Engineering, Vol. Intelligent Information Integration, Data & Knowledge Engineering, Vol. 36, Num. 1, Pages 215-249, Elsevier Science B.V. 2001 36, Num. 1, Pages 215-249, Elsevier Science B.V. 2001

D. Beneventano, S. Bergamaschi, F. Guerra, M. Vincini: "The MOMIS D. Beneventano, S. Bergamaschi, F. Guerra, M. Vincini: "The MOMIS approach to Information Integration", IEEE and AAAI International approach to Information Integration", IEEE and AAAI International Conference on Enterprise Information Systems (ICEIS01), Setúbal, Conference on Enterprise Information Systems (ICEIS01), Setúbal, Portugal, 7-10 July, 2001. Portugal, 7-10 July, 2001.

Silvana Castano, Valeria De Antonellis, Sabrina De Capitani di Silvana Castano, Valeria De Antonellis, Sabrina De Capitani di Vimercati: Global Viewing of Heterogeneous Data Sources. TKDE 13(2): Vimercati: Global Viewing of Heterogeneous Data Sources. TKDE 13(2): 277-297 (2001)277-297 (2001)


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

Components (actors and stages)Components (actors and stages)

There is another family of scenarios of interest, that is those concerning the creation of There is another family of scenarios of interest, that is those concerning the creation of a new node, the update of an existing node, and the cancellation of a node. These a new node, the update of an existing node, and the cancellation of a node. These scenarios describe the structural life of a SEWASIE system, namely its growth and scenarios describe the structural life of a SEWASIE system, namely its growth and change in time. change in time.

The The creation of a newcreation of a new node is the acquisition of new information sources and the node is the acquisition of new information sources and the organisation of them into an information unit (SINode). This is a semi-automatic organisation of them into an information unit (SINode). This is a semi-automatic process with the goal ofprocess with the goal of

configuring the appropriate wrappers allowing access to the data and their configuring the appropriate wrappers allowing access to the data and their structures,structures,

building an ontology, that is a global virtual view (GVV) and the mapping description building an ontology, that is a global virtual view (GVV) and the mapping description between the GVV itself and the integrated sources,between the GVV itself and the integrated sources,

configuring the query manager for optimal handling of queries within this node, andconfiguring the query manager for optimal handling of queries within this node, and notifying the brokering agents network about the new node (or instantiating a new notifying the brokering agents network about the new node (or instantiating a new

brokering agent for the new node)brokering agent for the new node) The The update of an existing nodeupdate of an existing node concerns structural changes within the node, i.e. concerns structural changes within the node, i.e.

changes of the ontology,changes of the ontology, changes of source structure which imply adaptation at the node level changes of source structure which imply adaptation at the node level addition/deletion of a source which imply a change of the ontology and adaptation at addition/deletion of a source which imply a change of the ontology and adaptation at

the brokering agent levelthe brokering agent level Notice that the above cited changes do not concern changes of the data content. Notice that the above cited changes do not concern changes of the data content. The The deletion of a nodedeletion of a node concerns the removal of the references to the node from all concerns the removal of the references to the node from all

the brokering agents in the network, and the subsequent termination of the activities the brokering agents in the network, and the subsequent termination of the activities of the node.of the node.


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

AgentsAgents

The SEWASIE project will develop a FIPA compliant trusted agent The SEWASIE project will develop a FIPA compliant trusted agent network, featuring completely open, scalable and secure-oriented network, featuring completely open, scalable and secure-oriented architecture issues with the aim of making available the knowledge as architecture issues with the aim of making available the knowledge as synthesized in semantically enriched nodes of a virtual network. synthesized in semantically enriched nodes of a virtual network.

The advantages of an agent architecture in a context like SEWASIE The advantages of an agent architecture in a context like SEWASIE are given byare given by

savings of bandwidth: the agents can move locally to the resources they savings of bandwidth: the agents can move locally to the resources they want to use and carrying along the code to manage themwant to use and carrying along the code to manage them

ability to deal with non-continuous network connections, and therefore be ability to deal with non-continuous network connections, and therefore be intrinsically suited for mobile computingintrinsically suited for mobile computing

On the other end, the use of mobile, autonomous agents may add On the other end, the use of mobile, autonomous agents may add some complexity to the overall picture, due to the potential autonomy some complexity to the overall picture, due to the potential autonomy and indeterminacy of their plans of action. and indeterminacy of their plans of action.


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

Query AgentsQuery Agents

A Query Agent is the actual carrier of a query from a user to the A Query Agent is the actual carrier of a query from a user to the system.system.

The term “query” is to be interpreted as a general statement in a The term “query” is to be interpreted as a general statement in a known intermediate query language which may be interpreted by known intermediate query language which may be interpreted by SINode components (query managers) within the system. This SINode components (query managers) within the system. This query includes information on the context of the user at the time query includes information on the context of the user at the time of the establishment of the query. of the establishment of the query.

This means that information about the specific activity of the user, This means that information about the specific activity of the user, his/her preferences, feedback on appreciation of the results of his/her preferences, feedback on appreciation of the results of similar queries in the past under similar circumstances, and so on, similar queries in the past under similar circumstances, and so on,

are embedded in the query.are embedded in the query.


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

Brokering AgentsBrokering Agents

The case of a STATIC WORLD, where one universal ontology with a reference vocabulary is defined beforehand, and all sources have to fit in there somehow; such a set-up may be envisioned for smaller, strongly structured worlds

The case of a DYNAMIC WORLD, where no universal ontology exists except as juxtaposition of all the existing ontological domains identified at any given time; this set-up appears to be typical of larger, open, partially structured, worlds with autonomous components

Moreover, it should be noted that the integration policy may be different at the SINode level and at the global level. In fact, a more stringent ontology at the SINode and a loose juxtaposition at the global level may be a reasonable starting point.

The brokering agents are responsible for maintaining the knowledge about the SEWASIE network and act as entry points for query agents from users. These agents may be deployed as entry points to SINodes, or as pure informants and therefore anywhere in the network.

A brokering agent – knows about the ontologies which are present in the underlying SINode, – has some information about related ontologies in other nodes, and– has generic information about other ontologies


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

Brokering AgentsBrokering Agents

The depth of the information of the BA becomes more and more shallow with the distance (with respect to some metrics) between the ontologies where it is “expert” (those of the underlying SINode) and other ontologies covered within the system. Its information on other (non local) ontologies is incomplete.

The brokering agent is able to meet a query agent and recognise that the query presented by the latter is within scope for its local ontologies. If the query presented by the query agent also matches the strains of ontologies which are known to the brokering agent as being present on other nodes, then the brokering agent will also direct the query agent towards such brokers for further processing. When the info comes back to the query agent from the local node, then the query agent may need to interact with the brokering agent to clarify the semantics and context of the result and possibly integrate it with the results from other nodes.

Whenever a match of the incoming query ontology does not occur with the local ontologies, then the brokering agent will provide routing information towards other brokers to the query agent, which will then leave and move to other nodes.

The second main functionality of the brokering agent is to receive and classify its local ontologies and the references to other ontologies in the system. This means that whenever a new node is born, or changes, or disappears, then the ontologies used within the node have to be published to the local broker, which will update its internal information and then broadcast to other brokers a manifesto of its available ontologies.

Reinforcements are possible for specific brokers when the incoming ontological info has a strong correlation with local ontologies; in this way specialist brokers may arise within the SEWASIE system, as well as pure informants on a topic or range of related topics.


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

Sewasie in a P2P architectureSewasie in a P2P architecture

P2P computing consists of an open-ended network of distributed P2P computing consists of an open-ended network of distributed computational peers, where each peer can exchange data and computational peers, where each peer can exchange data and services with a set of other peers called acquaintances.services with a set of other peers called acquaintances.In the general case, a P2P system has no centralized schema In the general case, a P2P system has no centralized schema and no central administration.and no central administration.

In the SEWASIE architecture, we rely on two centralized aspects:In the SEWASIE architecture, we rely on two centralized aspects: The brokering agent (global control) that holds the knowledge of the The brokering agent (global control) that holds the knowledge of the

overall networkoverall network The global schema or data repository of the networkThe global schema or data repository of the network

We can define two alternative P2P networks:We can define two alternative P2P networks: INTER SINode NetworkINTER SINode Network Brokering Agent Network Brokering Agent Network

[S. Bergamaschi, F. Guerra, [S. Bergamaschi, F. Guerra, Peer to Peer Paradigm for a Semantic Search Peer to Peer Paradigm for a Semantic Search EngineEngine, in proceedings of the , in proceedings of the International Workshop on Agents and Peer-to-International Workshop on Agents and Peer-to-Peer Computing, to appear in Peer Computing, to appear in LNCS 2530, SpringerLNCS 2530, Springer]]

(i.e. the case of a dynamic world)(i.e. the case of a dynamic world)


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s


The INTER SINode network allows all the SINodes The INTER SINode network allows all the SINodes to exchange informationto exchange information

A SINode provides to other SINodes the knowledge about A SINode provides to other SINodes the knowledge about the involved information sources.the involved information sources.

It is possible to specify coordination formulas that explain It is possible to specify coordination formulas that explain how the data in one peer must relate data in a how the data in one peer must relate data in a acquaintance.acquaintance.

The Brokering Agent NetworkThe Brokering Agent Network Within the Brokering Agent Network, each Brokering Agent Within the Brokering Agent Network, each Brokering Agent

communicates with other peers in order to have communicates with other peers in order to have information about the involved sources.information about the involved sources.


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

SEWASIE in a P2P architectureSEWASIE in a P2P architecture

SEWASIEInformation

Node

VirtualDataStore

Information Node

SEWASIEInformation

Node

VirtualDataStore

BrokeringAgent

SEWASIEBrokering

Agent

SEWASIEBrokering

Agent

Inter SINode Network

Brokering Agent Network


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s

Sewasie in a P2P architectureSewasie in a P2P architecture This architecture generates a distributed knowledge about This architecture generates a distributed knowledge about

the involved information sourcesthe involved information sources

The Brokering Agent P2P network may provide a support The Brokering Agent P2P network may provide a support for generating coordination formulas (e.g. by using schema for generating coordination formulas (e.g. by using schema matching, by deriving relations among the peers using matching, by deriving relations among the peers using inference techniques).inference techniques).

The Brokering Agent P2P network supports the generation The Brokering Agent P2P network supports the generation of the query plan in order to identify which are the SINodes of the query plan in order to identify which are the SINodes to be queried. In particular, the P2P Network can:to be queried. In particular, the P2P Network can:

Generate interest groups with nodes that have similar content.Generate interest groups with nodes that have similar content. Help the query optimization, by giving information about the “data Help the query optimization, by giving information about the “data

placement”. A peer knows how is distributed data and in this way placement”. A peer knows how is distributed data and in this way the query plan may take into account the existing resource and the query plan may take into account the existing resource and bandwidth constraints.bandwidth constraints.


SE

WA

SIE

– S

eman

tic

Web

s an

d A

gen

tS in

Inte

gra

ted

Eco

no

mie

s


SINode network is an alternative approach: we SINode network is an alternative approach: we maintain a single brokering agent, holding the maintain a single brokering agent, holding the knowledge of the network topology and we need knowledge of the network topology and we need a P2P layer in each SINode with the following a P2P layer in each SINode with the following functionalities:functionalities:

The P2P layer needs a protocol for establishing an The P2P layer needs a protocol for establishing an acquaintance dinamicallyacquaintance dinamically

The P2P layer offers semi-automated support for The P2P layer offers semi-automated support for generating coordination formulasgenerating coordination formulas

The P2P layer uses approaches for query processing of The P2P layer uses approaches for query processing of multi-database systems multi-database systems

The P2P layer should be able to advertise its ontologyThe P2P layer should be able to advertise its ontology

sewasie: a semantic search engine sonia bergamaschi, maurizio vincini università di modena e reggio...

Documents