Adaptando, Aprendendo e Integrando Modelos Semânticos.

Post on 06-Aug-2015

96 views 0 download

Transcript of Adaptando, Aprendendo e Integrando Modelos Semânticos.

Adaptando, Aprendendo e Integrando

Modelos Semânticos

Kate Revoredo

Department of Applied Informatics

Informatics Graduate Program

Federal University of the State of Rio de Janeiro (UNIRIO)

Primeiro Encontro em Computação Semântica@ UFRJ, Março 2015

ALICE research group

Adaptating,Adaptating,Learning andIntegratingConceptualizationsEnvironment

Overall Scope

Structured data

(Databases) Learning Ontologies

(OntoUML,

FOL, DL)

Business rules

(declare)servicesIntegrating

Documents

Social

Data sources

Adapting Business Process Models

(BPMN, XPDL, KIPN)

(declare)

Resources

Patterns

AntiPatterns

servicesSocial

media

Logs

(provenance data)

StoriesKnowledge base

Linguistic

(wordnet)

abcabcabcabc

Learning well-founded ontologies from texts

Structured data

(Databases) Learning Ontologies

(OntoUML,

FOL, DL)

Business rules

(declare)servicesIntegrating

Documents

Social

Data sources

Adapting Business Process Models

(BPMN, XPDL, KIPN)

(declare)

Resources

Patterns

AntiPatterns

servicesSocial

media

Logs

(provenance data)

StoriesKnowledge base

Linguistic

(wordnet)

abcabcabcabc

Motivation

•Wordnet is Semantically restricted•Expresses some relationships, but not philosophical meta-properties.

• “Does it have an identity principle?” (sortal)

• “Is certain property always required?” (Rigidity)

• “Does it depend on other things to exist?” (Relational Dependency)

•Useful information for Computational Linguistics•Useful information for Computational Linguistics•Understanding of the concept under analysis

•Example:• “Author executes submission of paper”

• Author depends on Paper

• Paper only exists if an author submits it

Author PaperExecutes submission

**

Challenges

How to…•identify the best meaning (sense) for a word given its context?

•determine the most adequate construct of the well-founded modeling

language to represent this sense?

Linguistic approach with semantic focus

WordNet(synsets database)

Semantic TypesUFO

(OntoUML)

Supersenses and

Semantic Relations

From Wordnet synsets to Semantic Types

•Simple Mappings•Simple correspondence between Supersense and Semantic Type

• Equal or similar names and definitions

•Plant � Flora•Plant: “plants”

Semantic TypesWordNet Synsets

7 / 40

•Plant: “plants”

•Flora: “things of the natural world which instances belong to the plant

kingdom”

•Exemplo: Tree, flower, grass

Semantically Expanding WordNet through Semantic Types and UFO

From Wordnet synsets to Semantic Types

•Complex Mappings•More information then definitions needed

• Hypernyms and Hyponyms, Holonyms and Meronyms, etc.• Syntactical derivation• Morfological characteristics of sinonyms.

Semantic TypesWordNet Synsets

8 / 40

• Morfological characteristics of sinonyms.

• It is necessary to use more than one Semantic Type to complete the meaning

•Artifact � Artefact and Parts•Artifact: “Man-made objects” (different from Object)•Artefact: “Things made by humans, i.e. not existing in nature”•Parts: “things that can be seen as parts of other things or beings ”

Semantically Expanding WordNet through Semantic Types and UFO

Method Proposal

9 / 21

Example of an ontology learned

Pattern-Based Ontology Refinement

Structured data

(Databases) Learning Ontologies

(OntoUML,

FOL, DL)

Business rules

(declare)servicesIntegrating

Documents

Social

media

Logs

Data sources

Adapting Business Process Models

(BPMN, XPDL, KIPN)

Resources

Patterns

AntiPatterns

Logs

(provenance data)

StoriesKnowledge base

Linguistic

(wordnet)

abcabcabcabc

Ontology Design Patterns

Proposal

Improving ontology alignment through

correspondence antipatterns

Structured data

(Databases) Learning Ontologies

(OntoUML,

FOL, DL)

Business rules

(declare)servicesIntegrating

Documents

Social

Data sources

Adapting Business Process Models

(BPMN, XPDL, KIPN)

(declare)

Resources

Patterns

AntiPatterns

servicesSocial

media

Logs

(provenance data)

StoriesKnowledge base

Linguistic

(wordnet)

abcabcabcabc

Proposal

15

Back to the overall scenario...

Structured data

(Databases) Learning Ontologies

(OntoUML,

FOL, DL)

Business rules

(declare)servicesIntegrating

Documents

Social

Data sources

Adapting Business Process Models

(BPMN, XPDL, KIPN)

(declare)

Resources

Patterns

AntiPatterns

servicesSocial

media

Logs

(provenance data)

StoriesKnowledge base

Linguistic

(wordnet)

abcabcabcabc

Some challenges...

•Probabilistic formalisms

•Big data

•Refinement in general.

•Incoherence alignments

•E-science Support Infrastructure•For learning, refinement and integration•For learning, refinement and integration

ThankThankThankThank youyouyouyou............

Adaptating,

Contact

katerevoredo@uniriotec.br

Adaptating,Learning andIntegratingConceptualizationsEnvironment