february 24, 2006 ontologies helena sofia pinto ( [email protected] )
TRANSCRIPT
February 24, 2006
Summary
• Knowledge sharing
• Ontologies– What is an ontology?
– Kinds of ontologies
– How are ontologies built?
• Kind of life cycle
• Ontology building processes
– Ontology building tools
– Application areas and challenges
– Where is the research?
February 24, 2006
Knowledge Sharing
• Problem:– The cost of knowledge based systems
– Building the knowledge base from scratch
• KB Components– Medical diagnosis and medical tutoring
• Vocabulary definition: disease, organ, pathogenic agents (bacteria, virus, etc), kinds of bacteria (coli, coccos – estreptococcos, estaphilococcos -, etc) etc. – ontology
– Electronic diagnosis vs medical diagnosis
• Raise hypothesis, test, refine, etc. – problem solving method
February 24, 2006
Knowledge Sharing
• Solution:– Reuse and Sharing of knowledge
• Translation of knowledge bases between different KR languages
• Arbitrary differences among systems belonging to the same family
• Remote access to the knowledge base of another system
• Meaning of what is shared: Lack of consensus about vocabulary
February 24, 2006
What is an ontology?
• Capture the static knowledge in a given domain that is accepted and sharable across applications and groups
• Defs:– An explicit formal specification of a shared conceptualization
– a vocabulary of terms and some specification of their meaning
February 24, 2006
What is an ontology?
• Set of symbols (concepts) + hierarchy (organized) + some specification of their meaning (restrict the possible interpretations for those symbols)
• Concepts are defined by their relations with other concepts
xptoxpt1
xpt2
O1
xvxv1
xv2
O2rel1
ist
February 24, 2006
What is an ontology?
• Distinction ontology/KB
– different role played by represented knowledge
• ontologies - k. +/- consensual of a community
» process, activity, resource
• kb - k. specific of a particular problem being solved, changes with inference
» activities of a particular enterprise; actual processes, activities, costs, resources used to build or produce a particular product; estimate of resources inferred to be needed to satisfy an order
February 24, 2006
What is an ontology?
• Depend on the application that powered its construction
– Same domain/ different tasks• a large number of common concepts• differently defined:
– different levels of detail (class, relation, etc.)– capturing different points of view (structural point of
view, functional point of view, etc.)– different levels of granularity
• There is no “The Ontology!” – genuine alternatives!!– Not to the philosophers
February 24, 2006
Kinds of ontologies
• representation or meta-– capture the representation primitives in a KR family or
paradigm (Frames: class, instance, relation -slots and facet-, function, etc.)
• general or upper– capture very general notions applicable across domains (Time:
time-point, time-range, duration, overlaps, before, after, etc.)
• domain– specific of a particular domain (Chemical elements: elements,
non-reactive elements, helium, non-metals, carbon, etc.)
• others...
February 24, 2006
Kinds of ontologies
D. McGuinness – Ontologies Come of Age
February 24, 2006
Origin
• An Ontology that describes the processes of the Central Dogma of Molecular Biology for prokaryotic organisms.
• Formalized and can be used by an inference engine to answer user questions
February 24, 2006
Origin: Processes
February 24, 2006
Origin Entities
February 24, 2006
Origin Relations (1)
February 24, 2006
Origin: Relations (2)
February 24, 2006
Origin: Roles
February 24, 2006
Origin: Transcription
February 24, 2006
Origin: Transcription, subactivities
February 24, 2006
Origin: Activities
February 24, 2006
How are they built?
• General Process
• Life cycle
• Sub-Processes:– from scratch
– by means of reuse:
• integration
• merge
February 24, 2006
General process
Especificação Conceptualização Formalização Implementação Manutenção
Aquisição de Conhecimento
Avaliação
Documentação
February 24, 2006
Life cycle
• Prototipização evolutiva
A1
Cascata Iterativo
A1A2 A2 A3A3
Evolutivo
A1 A2 A3
February 24, 2006
Methodologies to build from scratch
• There are a few methodologies to build ontologies from scratch
• None of existing methodologies from scratch is widely accepted
• It is still more of a craft than an engineering task
February 24, 2006
Methodologies to build from scratch
• Most representative methodologies are:
– TOVE methodology [Gruninger, Fox 1995]
– ENTERPRISE methodology [Uschold, King 1995]
– METHONTOLOGY [Fernández, Gómez-Pérez, Sierra 1999]
February 24, 2006
TOVE
TOVE activity corresponds to
Capture motivating scenarios and formulate informal competency questions
Specify terminology, formulate formal competency questions and specify axioms and definitions in FOL
Evaluate competency and completeness
Specification
Conceptualization, Formalization and Implementation
Evaluation
February 24, 2006
ENTERPRISE
ENTERPRISE activity corresponds to
Identify purpose and scope
Capturing knowledgeKnowledge Acquisition and Conceptualization
Specification
Evaluate Evaluation
CodingFormalization and Implementation
Document Documentation
February 24, 2006
Techniques
• Knowledge acquisition: – brainstorming, interviews, questionnaires, text analysis, mind
maps– experts, books, norms, etc.
• Conceptualization: – middle-out, grouping, glossary of terms, concept classification
trees
• Formalization:– intermediate tabular representations (concept dictionary, table
of binary relations, etc.)
February 24, 2006
Ontology evaluation
• One needs to guarantee quality
– technical evaluation: judge ontologies, their software environment and documentation against a framework:
• consistency, completeness, conciseness, etc.
– user assessment: judge from the user point of view the usability and usefulness of ontologies, their software environment and documentation when they are reused or shared in applications
• understandability, technically evaluated, portable, etc.
[Gómez-Pérez, Juristo, Pazos 1995] [Gómez-Pérez 1999]
February 24, 2006
Ontology evaluation
– ONTOCLEAN: analyze hierarchical taxonomy using philosophical principles.
• Aims:
– assure that instances do not violate class properties
– assure consistent hierarchical structure
[Guarino, Welty, 2001, 2002]
February 24, 2006
Ontology building tools
• Most important tools freely available:– PROTÉGÉ, http://protégé.stanford.edu/
– Ontolingua Server, http://WWW-KSL-SVC.stanford.edu:5915
– OntoEdit, http://ontoprise.de/products/ontoedit/
– KAON, http://kaon.semanticweb.org/
[Duineveld, Stoter, Weiden, Kenepa, Benjamins 1999]
Some provide help to identify similar concepts (merge):
PROTÉGÉ (PROMPT, ex-SMART)
Ontolingua Server (Chimaera)
February 24, 2006