OWSCIS: Ontology and Web Service based Cooperation of Information Sources

OWSCIS Ontology and Web Service based Cooperation of Information Sources SITIS'2007, Shanghai, China, December 2007 Raji Ghawi , Thibault Poulain, Guillermo Gomez and Nadine Cullot Laboratoire Electronique, Informatique et Image UMR CNRS 5158 Université de Bourgogne, Dijon, FRANCE

Outlines

Introduction

Related works

OWSCIS Architecture

Mapping Process

Querying Process

Decomposition

Resolution

Recomposition

Future Works

Introduction

Interoperability Problem

Distrubtion

Hetereogeniety

Instability

Solutions

Ontologies

Web Services

Related Works ( existing systems )

BUSTER

SIMS

KRAFT

COIN

Carnot

InfoSleuth

OBSERVER

…

Related Works (features)

Information Sources

DB, XML, files, ...

Dynamic .vs. Static

Architecture

Agent-based

Mediator-based

Use of Ontology

Single, Multiple, Hybrid

Ontology Specification Language

Query Process

1. OWSCIS O ntology and W eb S ervice - based C ooperation of I nformation S ources

Hybrid Ontology Approach

Local Ontologies and Reference Ontology

Web Service - Oriented Architecture

GAV (Global As View).

OWSCIS Architecture Knowledge Base Visualisation Web Service Data Provider Data Provider End User Data Provider Querying Web Service Mapping Web Service Local Ontology Reference Ontology Local Ontology Local Ontology

Knowledge Base Mapping Directory Reference Ontology Tool Box

Reference ontology

specific knowledge domaine

OWL-DL

Mapping directory

correspondances between terms of reference and local ontologies

Toolbox

tools and methods used by the mapping web service

Knowledge Base Visualisation Web Service Data Provider Data Provider End User Data Provider Querying Web Service Mapping Web Service Local Ontology Reference Ontology Local Ontology Local Ontology

Data Provider Knowledge Base Visualisation Web Service Data Provider Data Provider End User Data Provider Querying Web Service Mapping Web Service Local Ontology Reference Ontology Local Ontology Local Ontology

Information sources are wrapped to a local ontology

using DB2OWL for databases

Two types of mappings

Local ontology has no instances

DB2OWL Database Local Ontology Mappings Local Onto  Ref. Onto DB  Local Onto

Mapping Web Service Knowledge Base Visualisation Web Service Data Provider Data Provider End User Data Provider Querying Web Service Mapping Web Service Local Ontology Reference Ontology Local Ontology Local Ontology Concepts Similarity Roles Similarity Refining Mappings Estimation Mapping Web Service Mappings Directory Reference Ontology Tool Box Local Ontology

Mapping local ontologies to the reference domain ontology

Comparing ontologies using the methods defined in the toolbox

Storing produced mappings into the appropriate data provider

Updating the mappings directory

Querying Web Service Query Decomposition Query Recomposition Querying Web Service End User Knowledge Base Visualisation Web Service Data Provider Data Provider End User Data Provider Querying Web Service Mapping Web Service Local Ontology Reference Ontology Local Ontology Local Ontology Mappings Directory Reference Ontology Visualisation Web Service

Mapping users submit queries in terms of the reference ontology

A query is decomposed into a set of modular queries using the mapping directory

Each sub-query is resolved in a data provider

Partial results are recomposed giving the final query result

Visualisation Web Service Knowledge Base Visualisation Web Service Data Provider Data Provider End User Data Provider Querying Web Service Mapping Web Service Local Ontology Reference Ontology Local Ontology Local Ontology Visualisation Web Service End User Querying Web Service

Visualization of the reference ontology

Visualization of the queries and their results.

Enriching the results of the query using the semantic information described in the reference ontology.

Mapping Process

Preprocessing:

cleaning up the data

Similarity estimation:

giving a numerical similarity estimation value to all pairs of concepts

Semantic : extract known words from the concepts names and perform a semantic similarity estimation over them.

Structural : estimate structural similarity by comparing concept names as a string.

Refining:

solving cases where the similarity value between two concepts is neither high enough nor low enough to determine whether there is an equivalence or not.

Exploitation:

Translating similarities from their numerical values into mappings.

Automatically: producing the overall mapping between the two ontologies

Iteratively: the program suggests what appears to be the best mapping, and let the expert validate or not the choice.

Query Process: Overview Query Web Service recompose decompose Visualization Web Service Mapping Directory … SPARQL User query SPARQL subquery1 SPARQL result1 SQL query1 SQL Result1 Database Local Ontology DB ↔ Local Onto rewrite translate reformulate Ref. Onto ↔ Local Onto SQL statements DB2OWL Mapping Web Service Data Provider

Query Decomposition (Example) Global Ontology Local Ontology 1 Local Ontology 2 Person firstName lastName Student Author Session Reference title year Book ISBN Module module_name Hall hall_name building Publisher pub_name pub_address Article Lecturer book_author session_module session_hall book_publisher session_lecturer Person firstName lastName Student Session Module module.name Hall hall.name building Lecturer session.lecturer session.module session.hall Author author.lastName author.firstName Reference title year Book ISBN Article book.author

SELECT ?t WHERE { ?b ro:book_author ?a . ?b ro:title ?t . ?a ro:lastName ?ln . ?a ro:firstName ?fn . ?s ro:session_lecturer ?l . ?l ro:lastName ?ln . ?l ro:firstName ?fn . ?s ro:session_module ?m . ?m ro:module_name "Database" . } (d) Rewritten sub-query2 (c) Rewritten sub-query1 SELECT ?ln ?fn WHERE { ?s lo2:session.lecturer ?l . ?s lo2:session.module ?m . ?m lo2:module.name "Database". ?l lo2:lastName ?ln . ?l lo2:firstName ?fn . } SELECT ?ln ?fn ?t WHERE { ?b lo1:title ?t . ?b lo1:book.auther ?a . ?a lo1:author.lastName ?ln . ?a lo1:author.firstName ?fn . } (b) Sub-query2 (a) Sub-query1 SELECT ?ln ?fn WHERE { ?s ro:session_lecturer ?l . ?s ro:session_module ?m . ?m ro:module_name "Database" . ?l ro:lastName ?ln . ?l ro:firstName ?fn . } SELECT ?ln ?fn ?t WHERE { ?b ro:title ?t . ?b ro:book_author ?a . ?a ro:lastName ?ln . ?a ro:firstName ?fn . }

SPARQL-to-SQL translation BGP Construction SELECT ?ln ?fn WHERE { ?s lo2:session.lecturer ?l . ?s lo2:session.module ?m . ?m lo2:module.name "Database". ?l lo2:lastName ?ln . ?l lo2:firstName ?fn . } V0 V1 V2 V3 V4 SELECT module.moduleId AS c0, module.moduleName AS c1 FROM module lo2:module.name SELECT session.sessionId AS c0, module.moduleId AS c1 FROM session, module WHERE session.moduleId = module.moduleId lo2:session.module SELECT session.sessionId AS c0, lecturer.lecturerId AS c1 FROM session, lecturer WHERE session.lecturerId = lecturer.lecturerId lo2:session.lecturer SELECT person.personId AS C0, person.lastName AS C1 FROM person lo2:lastName SELECT person.personId AS C0, person.firstName AS C1 FROM person lo2:firstName ?s ?l ?m ?ln «Database» :session.lecturer :lastName :module.name :session.module ?fn :firstName

SPARQL-to-SQL translation Joining Statements SELECT V0.C1 AS ln, V1.C1 AS fn FROM ( SELECT person.personId AS C0, person.firstName AS C1 FROM person) AS V0, ( SELECT person.personId AS C0, person.lastName AS C1 FROM person) AS V1, ( SELECT session.sessionId AS C0, lecturer.lecturerId AS C1 FROM session, lecturer WHERE (session.lecturerId = lecturer.lecturerId)) AS V2, ( SELECT session.sessionId AS C0, module.moduleId AS C1 FROM session, module WHERE (session.moduleId = module.moduleId)) AS V3, ( SELECT module.moduleId AS C0, module.moduleName AS C1 FROM module) AS V4 WHERE (V0.C0 = V1.C0) AND (V2.C1 = V0.C0) AND (V2.C0 = V3.C0) AND (V3.C1 = V4.C0) AND (V4.C1 = 'Database') ?s ?l ?m ?ln «Database» :session.lecturer :lastName :module.name :session.module ?fn :firstName V0 V1 V2 V3 V4 C0 C1 C0 C1 C0 C1 C0 C1 C0 C1

Query Recomposition

Each sub-query result forms a new relation.

These relations are joined using shared variables.

R 1 (?ln,?fn,?t) R 2 (?ln,?fn) SELECT ?ln ?fn WHERE { ?s ro:session_lecturer ?l . ?s ro:session_module ?m . ?m ro:module_name "Database" . ?l ro:lastName ?ln . ?l ro:firstName ?fn . } SELECT ?ln ?fn ?t WHERE { ?b ro:title ?t . ?b ro:book_author ?a . ?a ro:lastName ?ln . ?a ro:firstName ?fn . } ?t  (<R1> <R2>) ?ln, ?fn

Implementation