Terminology model, Semantic Interfaces and Reference Model

1
Terminology model, Semantic Interfaces and
Reference Model
January 2003, R. Karge
Open Forum 2003 on Metadata RegistriesStatistics
Track January 23, 2003, 11.15 - 12.00
2
Introduction

• The terminology model, reference model and
  features
• Knowledge and metadata
• The core terminology model as expression of
  simplified knowledge
• The extended terminology model and missing
  features
• Using terminology model for metadata exchange and
  defining a reference model
• How terminology models can be used for building
  metadata systems and terminology model based
  tools

3
Why building terminology models

• Bridge the gap between statistical experts and IT
  experts
• Conceptual metadata definition by means of
  terminology models
• Formalize statistical expert knowledge for
  storing it in metadata systems
• Structuring knowledge
• Documentation of metadata models
• Harmonizing and exchange metadata between
  different organizations or systems

4
Metadata - Definition

• What is data
• A data item or a state is the representation of a
  fact described by object, time property and value
  (o, t, p, v).
• What is metadata
• There exists at least one abstract data state for
  each metadata state that is related to the
  metadata state.
• Metadata states are states that are not
  functional depending on data states, i.e.
  changing data states will not influence metadata.
  
• The set of data objects and the set of metadata
  objects is distinct (i.e. the objects described
  in the database and in the metadata database are
  distinct).
• Metadata describes properties (p), grouping and
  classification of objects (o) and time
  points/periods (t)
• Metadata can be presented as data and thus
  described by metadata (meta-metadata).

5
Basic principles in expressing knowledge

• Properties
• Defining properties and constraints (e.g. value
  domains)
• Grouping
• Grouping properties in types
• Classifying/grouping objects/time by means of
  concepts (defining groups of objects and time
  points/intervals)
• Rules
• Describing state transitions by means of
  rules/behavior
• Describing state operations by means of
  rules/behavior
• Causalities
• Describing causalities as dependent state
  transitions
• Describing state transitions as consequence of
  history

6
Terminology model (TM) requirements

• What the TM does
• The TM is a simplified and understandable
  presentation of knowledge.
• The TM describes the way knowledge is structured
  (properties and grouping)
• What the TM should do
• The TM should be able to describe rules
• What the TM not does
• The TM does not support causalities

7
Core level terminology model (Structure)

• Concept The concept defines a term that refers to
  a basic idea in a subject matter area. The names
  for concepts and synonyms must be unique in a
  terminology model. The concept is defined by
• Name The name is a single word or group of words
  that identifies the concept.
• Description A description or definition of the
  named concept
• Characteristic List of characteristics that
  describe the details of a concept.
  Characteristics are defined as characteristic.
• Synonyms List of synonyms that can be used
  instead of the concept name.
• Characteristic The characteristic of a concept
  defines a relevant detail (attribute) of a
  concept. The names and synonyms for
  characteristics must be unique within a concept
  definition. The characteristic is defined by
• Name Single word or group of words that
  identifies the characteristic
• Description The description or definition of the
  named characteristic.
• Related concept If the characteristic is not
  simply defined as text but refers to another
  concept the referenced concept is mentioned here.
  In documents referenced concepts are visualized
  as underlined terms.
• Synonyms List of synonyms defined for the
  characteristic.

8
Extended terminology model

• More information is needed to build UML models or
  semantic interfaces
• Additional information is needed to generate
  database models
• Modeling becomes more efficient when using
  concepts as generalization or reference
• The core TM support the grouping aspect of
  knowledge, but not the operational and causal
  aspects
• Some of the missing features are provided in the
  extended terminology model

9
Extended terminology model (concept)

• Concept The concept defines a term that refers to
  a basic idea in a subject matter area. The names
  for concepts and synonyms must be unique in a
  terminology model. The concept is defined by
• Name
• Description
• Characteristic
• Synonym
• Source Refers to the organization, individual or
  document that has originally defined the concept.
• Scope Defines the scope metadata objects of this
  type usually appear ( local global
  semi-global).
• Rules A list of one or more concept specific
  rules that describe the behavior of the concept.
  Rules are described as rule.
• Categories This is a list of categories in
  different classifications the concepts belongs
  to.
• Example One or more examples describing the
  defined concept.

10
Extended terminology model (characteristic)

• Characteristic The characteristic of a concept
  defines
• Name
• Description
• Related concept
• Synonym
• Type describes the way a characteristic is
  referenced as (generalization reference
  attribute).
• Value domain describes the possible values by
  means of a value domain.
• Source Refers to the organization, individual or
  document that has originally defined the
  characteristic.
• Rules A list of one or more characteristic
  specific rules that describe the behavior of the
  characteristic. Rules are described as rule.
• Edit level describes the importance of the
  characteristic (mandatory, optional, suggested)
• Cardinality describes the maximum number of
  elements that can be referenced as
  characteristic. Typical cases are 0,1 and N,
  where N refers to any number greater or equal to
  0.
• Inverse describes a characteristic in the related
  concept that refers as inverse reference to the
  described concept.
• Example One or more examples describing the
  defined characteristic.

11
Extended terminology model (rule)

• Rule A rule describes a specific behavior of a
  concept or characteristic. It may describe
  consistency rules as well as certain
  functionality associated with the concept or
  characteristic. The behavior described by the
  rule is typical for the concept or characteristic
  and not associated with a specific application.
  Thus, rules on this level describe business
  rules.
• Name Single word or group of words that
  identifies the rule.
• Description The description or definition of the
  rule.
• Source Refers to the organization, individual or
  document that has originally defined the rule.
• Example One or more examples describing the
  defined rule.

12
Extended terminology model (Value domain)

• Value Domain The value domain describes the
  permissible values for the characteristic.
• Name Single word or group of words that
  identifies the value domain.
• Description The description or definition of the
  value domain.
• Source Refers to the organization, individual or
  document that has originally defined the value
  domain.
• Value list describes the list permissible values
• Value The value describes a permissible values
  for a value domain.
• Name Single word or group of words that
  identifies the value.
• Description The description or definition of the
  value.
• Source Refers to the organization, individual or
  document that has originally defined the value.

13
Practical experiences Building TMs

• Practical work on different terminology models
  for statistical metadata
• 1999 the TM has been introduced as the base of
  the work of the Neuchatel group (classifications)
• 2000 Based on the TM the Oslo group started to
  define concepts for variables
• 2001 Registers, cubes and tables have been
  discussed in the SOS technical group based on a
  TM
• 2002 The Neuchatel group (Variables) was founded
  to continue defining variables and related
  concepts.
• Technologies have been provided to generate many
  different model presentations from a TM

14
Practical experiences Metadata harmonization

• Harmonization of metadata
• The core TM is a good base for harmonizing
  metadata models since it covers most common
  modeling principles (properties and types)
• Relational model (Relation, attributes)
• Object-oriented model (Object type, property)
• DTD (Element, elements/attributes)
• The core TM defines content and meaning that
  allows mapping different metadata models
• Mappings have been made for ComeIn, DDI and
  StatObject (in the frame of the METANET project).
  ISO 11179 is on the way.
• The TM allows describing the semantic of a
  metadata model and allows exchanging metadata on
  semantic level

15
Model Transformation

• Agreements required for interfaces
• Technical agreement (XML, COM, ...)
• Semantic agreement (semantic standard)
• Using the same semantic standard for different
  technical interfaces makes communication easier
• Metadata exchange is one example for a TM based
  communication interface

16
Model Transformation - Metadata Exchange via XML
Transfer Metadata
US BLF Metadata Repository
OFS Metadata Repository
17
The Reference Model (RM)

• The RM is a conceptual metadata model covering
  relevant metadata objects of a specific area
  (e.g. statistical metadata)
• The RM is a model that allows mapping different
  metadata models to each other
• The RM is independent on technical
  implementations
• The RM defines the relevant concepts and
  characteristics of all involved metadata models
• The RM follows basic rules for metadata modeling
• The RM is a base for harmonizing metadata
• The RM defines metadata exchange rules
• The RM is a quality measure for metadata models
• Classifying metadata objects in the RM helps
  mapping metadata models

18
The Reference Model - Requirements

• The reference model is defined as a two level
  model since this is the common denominator for
  most models
• Relational model (Relation, attributes)
• Object-oriented model (Object type, property)
• Terminology model (Concept, characteristics)
• DTD (Element, elements/attributes)
• The reference model must provide conceptual
  definitions because mapping is based on mapping
  concepts and not on terms.
• The reference model must be consistent, i.e. each
  referenced concept, object type or relation must
  be defined in the reference model.
• The reference model must not contain overlapping
  definitions, i.e. each concept or characteristic
  is defined only once in the reference model.
• A reference model provides a number of
  classifications. Each classification must define
  a number of distinct and complete categories on
  each level of the classification, i.e. each
  metadata object or concept must fit into exactly
  one of the lowest level categories within each
  defined classification.

19
The Reference Model - Levels
20
The Reference Model Object classifications
quality
operational
physical
Terms ...
Classifications ...
Variables and stat. objects...
Quality metadata ...
Statistical activities and processes ...

Administrative metadata ...
TM -based Tools
21
The Reference Model Object classifications
Thesaurus
Activities
Variables/Data Elements
Classification
Processes
TM -based Tools
22
Terminology model based tools
CORBA, COM, C ComeIn
Mapping rules
Business rules
Database access API
Database
23
More Information

• www.run-software.com
• www.run-software.com/ReferenceModel
• reinhard.karge_at_run-software.com

TM -based Tools