QNT Classification

1
QNT Classification

• - A New Approach to Knowledge Representation
• US and International patents pending
• Pavel Babikov, Oleg Gontcharov,
• and Maria Babikova
• QNT Software Development Inc.
• 528 Victoria Ave., Windsor, ON Canada N9A 4M8
• Tel [phone] Fax
  1-519-253-5389
• www.quasinewtonian.com

2
Requirements to Classification Scheme

• General polyhierarchical structure no global
  separation of classification aspects
• Persistence of the polyhierarchy no dependence
  on actual set of classified objects
• Compactness of description no explicit
  enumeration of categories
• Intrinsic support of set theory operations when
  forming classification categories
• Efficient realization of test for distance
  inheritance no combinatorial search
• Conceptual simplicity no application-specific
  program codes or cumbersome descriptions

3
Classification Criterion Atomic Specialization
Aspect
Criterion Ck
Ck Pk(i), i 1,2,...,Nk i branch
number Nk cardinality Pk(i) true / false
Pk(i) ? Pk(j) false for i ? j Elementary
attributes ak(i) Ck, i Pk(i) true
Criterion branches
4
Conjunctive Classificationsby Systems of Criteria
5
Trees and Facets - ClassicConjunctive Schemes
6
Common Disadvantages of Trees

• Mandatory ranking classification criteria the
  predefined path problem
• Massive duplication of criteria in parallel
  sub-trees the subtrees multiplication problem
• Purely conjunctive logical structure no
  intrinsic support for disjunctions and negations

7
Common Disadvantages of Facets

• Global separation of classification aspects
• The same predefined path and subtrees
  multiplication problems within facets
• Purely conjunctive intrinsic logical structure
  within facets
• Inelegant and cumbersome formalism supporting
  non-intrinsic features

8
Our Approach

• Generalize formalism for building classification
  in terms of elementary specializations by
  criteria
• Develop purely synthetic poly-hierarchical
  classification scheme based on partially ordered
  criteria

9

• Category can introduce more than one
  classification criterion
• Any applicable criterion can be used for further
  specialization

10

• Classification categories are described by
  logical formulae containing conjunctions,
  disjunctions and negations

11
Generating Polyhierarchy is Established by
Dependence Relationships Between Criteria
12
Generating Polyhierarchy of Criteria Implicitly
Defines Induced Polyhierarchy of Categories
13
A Fragment of Generating Polyhierarchy for
Classification of Means of Conveyance
More Complex Example
14
Advantages of the Method

• It satisfies all six requirements to
  classification scheme, listed above
• It provides very general and mathematically
  rigorous formalism for manipulating complex
  hierarchical information structures
• It uses very simple system of basic notions,
  without requiring special knowledge for
  implementation
• Target polyhierarchical classification is
  directly representable by DB structure no
  programming required!

15
Fields of Application

• Taxonomical, expert, logistic, and content
  management systems
• Enterprise resource planning and project
  management systems
• Application-specific data and knowledge bases
• Machine learning and AI systems
• Intelligent control systems and robots
• Electronic lists, catalogues and directories
• Internet search engines
• Online documentation and help subsystems
• Components of computer OS and compilers

16
Contact Information

• QNT Software Development Inc.
• 528 Victoria Ave.,
• Windsor, Ontario
• Canada N9A 4M8
• Tel [phone]
• Fax [phone]
• babikov_at_quasinewtonian.com
• gontcharov_at_quasinewtonian.com