Instituto de Engenharia de Sistemas e Computadores Investigação e Desenvolvimento em Lisboa...

Instituto de Engenharia de Sistemas e Computadores Investigação e Desenvolvimento em Lisboa

Ontology Building Process: The Wine Domain

João Graça, Márcio Mourão, Orlando Anunciação, Pedro Monteiro, H. Sofia Pinto, and Virgílio Loureiro

Instituto de Engenharia de Sistemas e Computadores Investigação e Desenvolvimento em Lisboa

Summary

• Context• Ontology• Wine Domain• Existing Wine Ontologies • Ontology Building Process

– Knowledge acquisition– Conceptualization– Evaluation– Formalization

• Future work

Instituto de Engenharia de Sistemas e Computadores Investigação e Desenvolvimento em Lisboa

Context

• Project for a Post-Graduate Course– Building an Ontology

• Team– 4 students from Artificial Intelligence area

• Theoretical knowledge• No prior practical experience

Instituto de Engenharia de Sistemas e Computadores Investigação e Desenvolvimento em Lisboa

Ontology

• What is an ontology– Concepts of a domain represented as:

• Classes - collection of individuals

• Instances – individuals

– Concepts are hierarchically organized• subclass, instance-of

– Other domain dependent relations• produced-in, producer, produced-by

Instituto de Engenharia de Sistemas e Computadores Investigação e Desenvolvimento em Lisboa

Wine Domain

• Type classified according to three main dimensions– Wine making process

• White, Rosé and Red

– Grape maturity state

– Fermentation process

WineAccordingMaturityState

Unripe Mature OverMature

WineAccordingFermentation

Sweet

Sparkling

Table

Instituto de Engenharia de Sistemas e Computadores Investigação e Desenvolvimento em Lisboa

Wine

• Classification– Wine dimensions

• Wine type (process, maturity, fermentation)

– Grape variety

– Harvesting year

– Producer

– Region

• Classification System– European: Defines requirements for a wine

– Rest of the World: Producer asserts the quality of the wine

• Characteristics– Measurable: alcohol percentage, tannin level and degree of acidity

– Ambiguous: vision, aroma and flavor

Instituto de Engenharia de Sistemas e Computadores Investigação e Desenvolvimento em Lisboa

Existing Wine Ontologies

• Noy and McGuinness, 2001– Mismatches between concepts and relations acquired and

represented• Ex: Dessert Wines

– Lack of restrictions• Red wine with white grapes is possible

• Conclusion– A new ontology to be built from scratch

Instituto de Engenharia de Sistemas e Computadores Investigação e Desenvolvimento em Lisboa

Ontology Building Process

• Knowledge Acquisition• Conceptualization• Evaluation• Formalization

Instituto de Engenharia de Sistemas e Computadores Investigação e Desenvolvimento em Lisboa

Knowledge Acquisition

• First Stage– Sources: Books and Internet

• Unstructured information

• Different definitions for the same concept

– Result• A glossary

• A general view about the domain

• Second Stage– Found a Wine Domain Expert

• Structured interviews

Instituto de Engenharia de Sistemas e Computadores Investigação e Desenvolvimento em Lisboa

Conceptualization (first stage)

• Conceptual model based on First Stage– Brainstorming into a Blackboard

– Attempt to create a model from concepts

– General Conceptual Model

Instituto de Engenharia de Sistemas e Computadores Investigação e Desenvolvimento em Lisboa

Conceptualization (second stage)

• First stage conceptual model evaluated by the Expert– Introduction of abstract concepts – Removal of very specific concepts– New simpler conceptual model

• Covers about 90% of the Domain

Instituto de Engenharia de Sistemas e Computadores Investigação e Desenvolvimento em Lisboa

Formalization

• OntoEdit– Proprietary tool– Beta Tester Program

• Concepts– Classes & Instances

• Relations between concepts– Attribute values of classes were restricted

• Impossible to create a red wine with only white grapes

• Impossible to assign a French region to a Portuguese wine

Instituto de Engenharia de Sistemas e Computadores Investigação e Desenvolvimento em Lisboa

Results

• OntoEdit has a built in Inference Engine: Ontobroker– Queries represented in F-Logic:– Auxiliar predicate:

FORALL Wine,Year, GrapeVariety hasGrapeVarietyYear(Wine,Year, GrapeVariety) <- Wine:#Wine[#hasYear»Year], Wine[#hasGrapeVariety‑» GrapeVariety].

- Query – Select all grapes that were used in wines from the year 2001FORALL Wine,GrapeVariety <- hasGrapeVarietyYear (Vinho,2001,Casta).

Wine GrapeVariety

Quinta de Roriz Tinto Reserva 2001 Touriga Nacional

Quinta de Roriz Tinto Reserva 2001 Tinta Roriz

Quinta do Cotto Tinto 2001 Touriga Nacional

Instituto de Engenharia de Sistemas e Computadores Investigação e Desenvolvimento em Lisboa

Future Work

• Refine the ontology

• Wine advisory system– Dialogue system

• Increase the knowledge base