Transcript Formal ontologies

Issues in Ontology-based
Information integration
By Zhan Cui, Dean Jones and Paul O’Brien
E-commerce
Requirement:
• Retrieve and Integrate information from
multiple resources.
• Details of the resources are hided from
users.
Obstacles:
• Understand queries.
• Determine resources.
• Integrate information.
Four levels in interoperability problems
1. System level: incompatible hardware
and operating system.
2. Syntactic level: different language and
data representations.
3. Structural level: different data models.
4. Semantic level: the meaning of terms.
e.g., synonyms.
Many technologies address problems in
first three levels, such as CORBA, DCOM, etc.
XML vs. Semantic Heterogeneity Problems
XML:
• Can solve problems in schema level.
• Provide common syntax for exchanging
heterogeneity information.
• Some standards for e-commerce:
e.g., ebXML.
XML:
• Cannot solve problems in semantic level.
• Terminology may not consistent in one file
or in a set of files.
Solution
• Formally specify the meaning of the
terminology of each system
(Formal ontology)
• Define a translation between each
system terminologies and an
intermediate terminology.
(Ontology mapping)
Outline
• Issues in Resolving Semantic
Heterogeneity
• Description of DOME
• DOME Demonstrator
• Conclusion
Issues in Resolving Semantic Heterogeneity
--- Developing ontologies
Formal ontology:
• A formal ontology consists of definitions of terms.
• It usually includes concepts with associated attributes,
relationships and constraints defined between the concepts and
instances of concepts.
Formal ontologies include different type of ontologies for different
purposes:
• Resource ontologies: define the terminology used by specific
information resources.
• Personal ontologies: define the terminology of a user or some
group of users.
• Shared ontologies: the common terminology between a number
of different systems.
Issues in Resolving Semantic Heterogeneity
--- Developing ontologies
The best approach to develop ontologies is usually determined
by the eventual purpose of the ontologies.
For example:
•Resource ontologies: bottom-up approach.
•Shared ontologies: top-bottom approach.
Issues in Resolving Semantic Heterogeneity
--- Mapping Between Ontologies
•Human intervention is necessary.
•Some tools are helpful: mediator systems, mapping libraries
and conversion functions.
•Mapping is not accurate. Information could be lost. This is
unacceptable for e-commerce.
Issues in Resolving Semantic Heterogeneity
--- Ontologies and Resource Information
• How to choose resources?
• It is necessary for resources to describe themselves:
resource ontologies.
• Personal ontologies are important for the system to
understand queries exactly.
• Many issues in locating resources:
e.g., users prefer one resource over another;
Issues in Resolving Semantic Heterogeneity
--- Ontologies and Database Schemas
Schema vs. Ontology
The main difference is their purposes.
• A schema is developed in order to model some data.
• A ontology is developed to define the meaning of the terms.
A resource has a formal ontology. Data are store in
database based on schema. Mapping between formal
ontology and resource schema is necessary.
Issues in Resolving Semantic Heterogeneity
--- Entity Correspondence
• There may be a lot of resources related to one query.
• Information have to be integrated to answer query.
• Construct correspondence between entities across
resources.
• Key attributes can be used to build correspondence.
• It is hard to determine whether information from different
resources is same or not.
DOEM Overview
DOEM (Domain ontology Management Environment)
• Ontology-based techniques.
• Designed for data reuse and knowledge
sharing.
• Retrieve information from multiple
resources to answer queries.
• Present results in a consistent way.
The DOEM architecture
DOEM Overview
--- Engineering client
• Develop and administrate a DOEM system.
• Extract (semi-automated) ontologies from legacy
system to define ontologies and mappings.
• Allow engineers to select best developing approach:
top-down or bottom-up.
• Engineers: define mapping between resource
ontologies and shared ontologies, resources and
shared ontologies, database schemas and resource
ontologies.
DOEM Overview
--- Ontology server
• Store ontologies defined using the engineering client.
• Allow user to access: share ontologies, resource
ontologies, application ontologies.
• Access through OKBC interface.
• Implement ontologies using the description logic
CLASSIC which can store ontologies and make
inference.
DOEM Overview
--- User client
• Interface to access system.
• Query information space.
• Load and browse ontologies.
• Queries and results use the same terminology.
DOEM Overview
--- Mapping server
• Store mappings between ontologies.
• Store generic conversion functions.
• Use a declarative syntax.
• Can be queried by query engine.
DOEM Overview
--- Wrappers
• Most interaction between a resource and the
DOME network occurs via wrappers.
• Translate queries between DOME and resources.
• Translate information that will be put into the
terminology of the particular resource.
DOEM Overview
--- Resource Directory
• Let system know which resources are available and
what these resources are.
• Store the directories and descriptions of resources.
DOEM Overview
--- Query engine
• Obtain a list of currently available and
relevant resources from resource
directory.
• Decompose the query into sub-queries.
• Send the sub-queries to the resources.
• Translate queries from the ontology of the
query to that of the relevant resource.
• Integrate results.
The DOEM architecture
DOME Demonstrator
• Based on a database of marketing
scenario.
• DOME controls mapping and limits
resources.
Conclusions
DOME:
• Solve information query at semantic level with formal
ontologies and ontology mappings.
• Provide an integrated view of networked
heterogeneous databases.
• Allow a user to select and browse definitions of
terminologies.
Comments
• General description.
• There is no details and experiments.
• No new technique is introduced.