19
Building an Operational Product Ontology System Written by Taehee Lee, Ig-hoon Lee, Suekyung Lee, Sang-goo Lee (IDS Lab. SNU) Dongkyu Kim, Jonghoon Chun (Prompt) Hyunja Lee, Junho Shim (SWU) ELSESEVIER, Electronic Commerce Research and Applications 5 (2006) 16–28 Presented by Dongjoo Lee IDS Lab., CSE, SNU

Building an Operational Product Ontology System Written by Taehee Lee, Ig-hoon Lee, Suekyung Lee, Sang-goo Lee (IDS Lab. SNU) Dongkyu Kim, Jonghoon Chun

Embed Size (px)

Citation preview

Page 1: Building an Operational Product Ontology System Written by Taehee Lee, Ig-hoon Lee, Suekyung Lee, Sang-goo Lee (IDS Lab. SNU) Dongkyu Kim, Jonghoon Chun

Building an Operational Product Ontology System

Written by Taehee Lee, Ig-hoon Lee, Suekyung Lee, Sang-goo Lee (IDS Lab. SNU)

Dongkyu Kim, Jonghoon Chun (Prompt)

Hyunja Lee, Junho Shim (SWU)

ELSESEVIER, Electronic Commerce Research and Applications 5 (2006) 16–28

Presented by Dongjoo Lee

IDS Lab., CSE, SNU

Page 2: Building an Operational Product Ontology System Written by Taehee Lee, Ig-hoon Lee, Suekyung Lee, Sang-goo Lee (IDS Lab. SNU) Dongkyu Kim, Jonghoon Chun

Copyright 2008 by CEBT 2

Ontology Creation

Creating ontology for a domain gives chances to

Analyze domain knowledge

Make domain assumptions explicit

Separate domain knowledge from operational knowledge

Provide common understanding of the information structure

Enable reuse of domain knowledge

Created domain ontology can be used for

Searching, browsing, integration, and configuration

Page 3: Building an Operational Product Ontology System Written by Taehee Lee, Ig-hoon Lee, Suekyung Lee, Sang-goo Lee (IDS Lab. SNU) Dongkyu Kim, Jonghoon Chun

Copyright 2008 by CEBT 3

Product Ontology

Product information is an essential component in e-commerce.

Distributed business data integration

Supply chain management

Spend analysis

E-procurement

Public Procurement Services (PPS) of Korea

G2B e-procurement service

Built in September 2002, 90% G2B transactions

KOCIS: Ontology based e-catalog System

http://www.g2b.go.kr:8100/index.jsp

Page 4: Building an Operational Product Ontology System Written by Taehee Lee, Ig-hoon Lee, Suekyung Lee, Sang-goo Lee (IDS Lab. SNU) Dongkyu Kim, Jonghoon Chun

Copyright 2008 by CEBT 4

Participants of KOCIS

Page 5: Building an Operational Product Ontology System Written by Taehee Lee, Ig-hoon Lee, Suekyung Lee, Sang-goo Lee (IDS Lab. SNU) Dongkyu Kim, Jonghoon Chun

Copyright 2008 by CEBT 5

Building Product Ontology

Modeling

Ontology Subsystems

Construction and maintenance

Search

Page 6: Building an Operational Product Ontology System Written by Taehee Lee, Ig-hoon Lee, Suekyung Lee, Sang-goo Lee (IDS Lab. SNU) Dongkyu Kim, Jonghoon Chun

Copyright 2008 by CEBT 6

Models – meta modeling

A meta-model is yet another abstraction and highlighting properties of the model itself

3-level meta modeling

M0 meta-class level

– Products, classification schemes, attributes, Unit Of Measures (UOMs)

– Meta relationships

M1 class level

– a snapshot or instance of the product ontology model in M0

M2 instance level

– Physical ontology data managed by the system

Page 7: Building an Operational Product Ontology System Written by Taehee Lee, Ig-hoon Lee, Suekyung Lee, Sang-goo Lee (IDS Lab. SNU) Dongkyu Kim, Jonghoon Chun

Copyright 2008 by CEBT 7

M0: Meta-class level

Page 8: Building an Operational Product Ontology System Written by Taehee Lee, Ig-hoon Lee, Suekyung Lee, Sang-goo Lee (IDS Lab. SNU) Dongkyu Kim, Jonghoon Chun

Copyright 2008 by CEBT 8

M1: Class level

Page 9: Building an Operational Product Ontology System Written by Taehee Lee, Ig-hoon Lee, Suekyung Lee, Sang-goo Lee (IDS Lab. SNU) Dongkyu Kim, Jonghoon Chun

Copyright 2008 by CEBT 9

M2: Implementation

Modeling goal is not only to design a conceptual product ontology model but also to implement it as an opera-tional ontology database model.

Through what?

OWL or RDFS?

– General purpose reasoning capability

– No robust OWL engine to practically handle a large knowledge-base

RDBMS?

– Restricted reasoning capability

– Shows high performance for low level semantic operations– Implement ontology subsystem to provide just enough reason-

ing capabilities along the core concepts

Page 10: Building an Operational Product Ontology System Written by Taehee Lee, Ig-hoon Lee, Suekyung Lee, Sang-goo Lee (IDS Lab. SNU) Dongkyu Kim, Jonghoon Chun

Copyright 2008 by CEBT 10

class

Attr

class

class

value

UOM

AttrUOM

Synonym

Attr

UOM

valueAttr

class value

UOM

Reasoning Capabilities through Technical Dictionary

VocVoc

Search

Mapping

PropertyHierarchy

Instance

PropertyConstraint

Constraint

Conversion

Instance

Synonym

InstanceInstance

InferencesLv1 Inference

Attr

UOMvalue

class

Attr

class

value

classAttr UOM

value

UOM

class

class

class

value

class

AttrPropertyclass

Attr

class

class

value

UOM

AttrUOM

Synonym

Attr

UOM

valueAttr

class value

UOM

VocVoc

Search

Mapping

PropertyHierarchy

Instance

PropertyConstraint

Constraint

Conversion

Instance

Synonym

InstanceInstance

Attr

UOMvalue

class

Attr

class

value

classAttr UOM

value

UOM

class

class

class

value

class

AttrProperty

LCD

LCD PANEL

classAttr

TD1 Class & Relationships

TD2 Product Attributes

TD3 UOMs

TD4 Product Values

TD5 Vocabularies

TD6 Class-Product relations

TD7 Class-Attribute relations

TD8 Attribute-UOM relations

TD9 Vocabulary relations

eOTD, GDD, RNTD, EC-CMA, EAN/UCC, RosettaNet, …

Page 11: Building an Operational Product Ontology System Written by Taehee Lee, Ig-hoon Lee, Suekyung Lee, Sang-goo Lee (IDS Lab. SNU) Dongkyu Kim, Jonghoon Chun

Copyright 2008 by CEBT 11

G2B classification TD

Page 12: Building an Operational Product Ontology System Written by Taehee Lee, Ig-hoon Lee, Suekyung Lee, Sang-goo Lee (IDS Lab. SNU) Dongkyu Kim, Jonghoon Chun

Copyright 2008 by CEBT 12

Ontology Subsystems

WAS

Legacy System

Legacy DB

XML

온톨로지 애플리케이션 서버Construction Search

Maintenance

Synchronizer TD Manager Model Manger Log Manger

DB ManagerCategory Manager Miner

Loader

Analyzer

Distributer

Searcher

Parser

Infer Manager

Ranker

Catalog Builder

XML Publisher

XML/Excel ConverterCategory Mapper

Ontology Database

Attr Product Voc-Rel Class-Attr

Class-ProdVocClass UOM Attr-UOM

Ontology System

RMI Communication

Page 13: Building an Operational Product Ontology System Written by Taehee Lee, Ig-hoon Lee, Suekyung Lee, Sang-goo Lee (IDS Lab. SNU) Dongkyu Kim, Jonghoon Chun

Copyright 2008 by CEBT 13

Probabilistic Similarity Computa-tion

Page 14: Building an Operational Product Ontology System Written by Taehee Lee, Ig-hoon Lee, Suekyung Lee, Sang-goo Lee (IDS Lab. SNU) Dongkyu Kim, Jonghoon Chun

Copyright 2008 by CEBT 14

Probabilistic Similarity Computa-tion

Page 15: Building an Operational Product Ontology System Written by Taehee Lee, Ig-hoon Lee, Suekyung Lee, Sang-goo Lee (IDS Lab. SNU) Dongkyu Kim, Jonghoon Chun

Copyright 2008 by CEBT 15

Visualization

Page 16: Building an Operational Product Ontology System Written by Taehee Lee, Ig-hoon Lee, Suekyung Lee, Sang-goo Lee (IDS Lab. SNU) Dongkyu Kim, Jonghoon Chun

Copyright 2008 by CEBT 16

Conclusion

Developed a practical product ontology system.

Product ontology database

Ontology subsystems.

– Construction and maintenance

– Search

Based on Bayesian belief network

Meta-modeling

Concepts: Products, classification schemes, attributes, and UOMs

Relationships

Functions

Standard reference system for e-catalog construction

Supply tools and operations for managing catalog standards

Knowledge base

– Design and construction of product database

– Search and discovery of products and services

Page 17: Building an Operational Product Ontology System Written by Taehee Lee, Ig-hoon Lee, Suekyung Lee, Sang-goo Lee (IDS Lab. SNU) Dongkyu Kim, Jonghoon Chun

Copyright 2008 by CEBT 17

Discussion

Uncovered semantics for handling inconsistencies

Constraints: domain, range, and cardinality

– foreign key constraints for ObjectTypeProperty

– data type constraints for DataTypeProperty

Triggers

OWL(RDF) export capability

Modeling based on OWL constructor

Generating schema and instances from rdbms

Querying performance comparison of RDF storages

Page 18: Building an Operational Product Ontology System Written by Taehee Lee, Ig-hoon Lee, Suekyung Lee, Sang-goo Lee (IDS Lab. SNU) Dongkyu Kim, Jonghoon Chun

Copyright 2008 by CEBT

Model based on OWL

18

ec:G2BCategory

ec:G2B[XX]

rdfs:subClassOf

ec:PRO[XX]

rdf:type

owl:Class

rdf:type

ec:GUNGBCategory ec:UNSPSCCategory

ec:GUNGB[XX]

rdfs:subClassOf

rdf:typeec:belongsTo

ec:belongsTo

ec:UNSPSCCategory

ec:belongsTo

ec:belongsTo

ec:UOM

rdf:typerdf:type

rdf:type

ec:UG[XX]

rdfs:subClassOf

rdfs:subClassOf

ec:UOM[XX]

rdf:type

rdf:type

ec:Quantity

#unnamed

rdf:type

rdf:type

ec:hasUOM

xml:string

ec:hasName

ec:productProperty

ec:has[XX]

ec:hasAG[XX]

rdfs:subPropertyOf

rdfs:subPropertyOf

owl:ObjectPropertyrdf:type

owl:TransitiveProperty

rdf:type

ec:hasProductValue

rdf:typerdfs:subPropertyOf

ec:Product

rdf:type

rdf:type

ec:valueProperty

rdf:type

Complexity: OWL-DL

ec:ProductValue

owl:unionOf

Page 19: Building an Operational Product Ontology System Written by Taehee Lee, Ig-hoon Lee, Suekyung Lee, Sang-goo Lee (IDS Lab. SNU) Dongkyu Kim, Jonghoon Chun

Copyright 2008 by CEBT 19

Querying Performance Compari-son

Simple queries

Complex queries that require inference

From 2007 MS thesis of Yucheon Lee.