Download pptx

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Concurrency control wikipedia , lookup

Relational model wikipedia , lookup

Database model wikipedia , lookup

Transcript
ONTOLOGY MATCHING
Part III: Systems and evaluation
6. Overview of matching systems
• 1. Schema-level information
• 2. Instance-level information
• 3. Both schema-level and instance-level information
• 4. overview meta-matching system
6.1 Schema-based systems
• 6.1.1 DELTA(Data Element Tool-based Analysis)
• discover attributes correspondences among database schemas
• relational schemas and extended entity-relationship(EER)
• use textual similarities
• returns a ranked list of documents
• 6.1.2 Hovy
• heuristics used to match large-scale ontologies
• Three types of matchers:
• concept names
• concept definitions
• Taxonomy structure
• the combined scores in descending order
6.1 Schema-based systems
• 6.1.3 TransScm
• provides data translation and conversion mechanisms
• by using rules, alignment is produced
• this alignment is used to translate data instances
• 6.1.4 DIKE(Database Intentional Knowledge Extractor)
• supporting construction of cooperative information(CISs)
• takes a set of databases belonging to the CIS
• Builds a kind of mediated schema
• 6.1.5 SKAT and ONION(Semantic Knowledge Articulation Tool)
• discovers mappings between two ontologies
• input ontologies -> graphs
• rules -> in first order logic
• ONION is successor system to SKAT
6.1 Schema-based systems
• 6.1.6 Artemis(Analysis of Requirements: Tool Environment for
Multiple Information Systems)
• a module of the MOMIS
• performs affinity-based analysis and hierarchical clustering of
database schema elements
• 6.1.7 H-Match
• ontology matching system
• for open networked systems
• inputs two ontologies and output correspondences
• 6.1.8 Tess(Type Evolution Software System)
• support schema evolution by matching the old and the new
versions
• Schemas are viewed as collection of types
• Matching is viewed as generation of derivation rules
6.1 Schema-based systems
• 6.1.9 Anchor-Prompt
• formerly known as SMART
• ontology merging and alignment tool
• Sequential matching algorithm that takes as input two ontologies
• handles OWL and RDF schema
• 6.1.10 OntoBuilder
• information seeking on the web
• operates in two phases:
• ontology creation(the training phase)
• ontology adaptation(the adaptation phase)
• 6.1.11 Cupid
• implements an algorithm comprising linguistic and structural
schema matching techniques
• computing similarity coefficients
6.1 Schema-based systems
• 6.1.12 COMA and COMA++(COmbination of MAtching algorithms)
• schema matching tool based on parallel composition of matchers
• provides:
• extensible library of matching algorithms
• a framework for combining obtained results
• platform for the evolution of the effectiveness
• 6.1.13 Similarity flooding
• is based on the idea of similarity propagation
• Schemas are presented as directed labeled graphs
• 6.1.14 XClust
• tool for integrating multiple DTDs
• based on clustering
6.1 Schema-based systems
• 6.1.15 ToMAS(Toronto Mapping Adaptation System)
• automatically detects and adapts mappings
• assumed:
• the matching step has already been performed
• correspondences have already been made operational
• 6.1.16 MapOnto
• constructing complex mappings
• inputs:
• an ontology specified in an ontology representation language(OWL)
• relational or XML schema
• simple correspondences between XML attributes and ontology datatype
properties
6.1 Schema-based systems
• 6.1.17 OntoMerge
• ontology translation on the semantic web
• dataset translation
• generating ontology extensions
• query answering from multiple ontologeis
• perform ontology translation by ontology merging and automated
reasoning
• 6.1.18 CtxMatch and CtxMatch2
• uses a semantic matching approach
• translates the ontology matching problem into the logical validity
problem
6.1 Schema-based systems
• 6.1.19 S-Match
• the first version rationalized re-implementation of CtxMatch with a few
added functionalities
• evolutions
• limited to tree-like structures
• 6.1.20 HCONE
• domain ontology matching and merging
• first, an alignment between two input ontologies is computed
• then, the alignment is processed
• 6.1.21 MoA
• ontology merging and alignment tool
• consists of:
• Library of methods for importing, matching, modifying, merging ontologies
• Shell for using those methods
• based on concept (dis)similarity derived from linguistic clue
6.1 Schema-based systems
• 6.1.22 ASCO
• discovers pairs of corresponding elements in two input ontologies
• handles ontologies in RDF Schema and computes alignments
between classes, relations, and classes and relations
• new version, ASCO2, deals with OWL ontologies
• 6.1.23 BayesOWL and BN mapping
• probabilistic framework
• includes the Bayesian Network mapping module
• in three steps:
• two input ontologies are translated into two Bayesian networks
• matching candidates are generated between two Bayesian networks
• concepts of the second ontology are classified with respect to the
concepts of the first ontology
6.1 Schema-based systems
• 6.1.24 OMEN(Ontology Mapping ENhancer)
• probabilistic ontology matching system based on Bayesian network
• inputs: two ontologies and initial probability distribution derived
• returns: a structure level matching algorithm
• 6.1.25 DCM framework
• a middleware system
• inputs: multiple schemas
• returns: alignment between all of them
6.2 Instance-based systems
• 6.2.1 T-tree
• an environment for generating taxonomies and classes from
objects(instances)
• Infer dependencies between classes(bridges) of different
ontologies
• input: a set of source taxonomies(viewpoints) and a destination
viewpoint
• returns: all the bridges in a minimal fashion
• 6.2.2 CAIMAN
• a system for document exchange
• Calculate a probability measure between the concepts of two
ontologies
6.2 Instance-based systems
• 6.2.3 FCA-merge
• uses formal concept analysis techniques
• tree steps:
• Instance extraction
• concept lattice computation
• Interactive generation of the final merged ontology
• 6.2.4 LSD(Learning Source Descriptions)
• discovers one-to-one alignments between the elements of source
schemas and a mediated schema
• learn from the mappings created manually between the mediated
schema and some of the source schemas
6.2 Instance-based systems
• 6.2.5 GLUE
• a successor of LSD
• employs mulitple machine learning techiques
• joint distributions of the classes
• 6.2.6 iMAP
• discovers one-to-one(amount ≡ quantity) and complex(address ≡
concat(city, street)) mapping between relational database
schemas.
• uses multiple basic matchers(searches)
6.2 Instance-based systems
• 6.2.7 Automatch
• mappings between the attributes of database schemas
• assumption:
• several schemas from the domain under consideration have already
been manually matched by domain experts
• 6.2.8 SBI&NB
• SBI(Similarity-Based Integration)
• SBI&NB is extension of SBI
• Determine correspondences between classes of two classifications
by statistically comparing the memberships of the documents to
these classes
6.2 Instance-based systems
• 6.2.9 Kang and Naughton
• a structural instance-based approach
• two table instances are taken as input
• 6.2.10 Dumas(DUplicate-based MAtching of Schemas)
• identifies one-to-one alignment between attributes by analyzing the
duplicates in data instances of the relational schemas
• looks for similar rows or tuples
• 6.2.11 Wang and colleagues
• one-to-one alignments among the web databases
• presents a combined schema model
• Global-interface, global-result, interface-result, interface-interface, and
result-result
6.2 Instance-based systems
• 6.2.12 sPLMap(probabilistic, logic-based mapping between
schemas)
• framework that combines logics with probability theory
6.3 Mixed, schema-based and instancebased systems
• 6.3.1 SEMINT(SEMantic INTegrator)
• a tool based on neural networks
• supports access to a variety of database system
• extracts from two databases all the necessary information
• using a neural network as a classifier
• 6.3.2 Clio
• managing and facilitating data transformation and integration
• focused on making the alignment operational
• transforms the input schemas into an internal representation
• taking the value correspondences(the alignment) together with
constraints coming form the input schema
6.3 Mixed, schema-based and instancebased systems
• 6.3.3 IF-Map(Information-Flow-based Map)
• based on the Barwise-Seligman theory of information flow
• matches two local ontologies by looking at how these are related to
a common reference ontology
• 6.3.4 NOM(Naïve Ontology Mapping) and QOM(Quick Ontology
Mapping)
• NOM adopts parallel composition of matchers from COMA
• QOM is a variation of the NOM
• QOM produces correspondences fast
6.3 Mixed, schema-based and instancebased systems
• 6.3.5 oMap
• a system for matching OWL ontologies
• built on top of the Alignment API
• uses several matchers(classifiers)
• 6.3.6 Xu and Embley
• proposed composition approach to discover one-to-one alignments,
onto-to-many and many-to-many correspondences between graphlike structures
• matches by combination of multiple matchers and with the help of
external knowledge recourses
6.3 Mixed, schema-based and instancebased systems
• 6.3.7 Wise-Integrator
• performs automatic integration of Web Interfaces of Search
Engines
• unified interface to e-commerce search engines of the same
domain of interest
• Attribute matching based on two types of matches: positive and
predictive
• 6.3.8 OLA(OWL Lite Aligner)
• balancing the contribution of each of the components that compose
an ontology
• inputs:OWL
6.3 Mixed, schema-based and instancebased systems
• 6.3.9 Falcon-AO
• a system for matching OWL ontologies
• components: those for performing linguistic and structure matching
• LMO is a linguistic matcher
• GMO is a bipartite graph matcher
• 6.3.10 RiMOM(Risk Minimization based Ontology Mapping)
• inspired by Bayesian decision theory
• inputs: two ontologies
• Aims at an optimal and automatic discovery of alignments which
can be complex
6.3 Mixed, schema-based and instancebased systems
• 6.3.11 Corpus-based matching
• Besides input information available from schema under
consideration
6.4 Meta-matching systems
• 6.4.1 APFEL(Alignment Process Features Estimation and
Learning)
• A machine learning approach that explores user validation of initial
alignments for optimizing automatically the configuration
parameters of some of the matching strategies of the system
• 6.4.2 eTuner
• Models:
• L is library of matching components
• G is a directed graph which encodes
• K is a set of knobs to be set