Intelligent Systems Development

1
Intelligent Systems Development

• Knowledge Engineering

2
Tatiana Gavrilova
Prof. Tatiana Gavrilova DSc, Saint-Petersburg
State Polytechnical University, Intelligent
Computer Technologies Dpt., Computer Science
School Consulting company Balt-Audit-Expert,
E-mail gavr_at_bae.ru, www.csa.ru/ailab
3
Saint-Petersburg State Technical
UniversityIntelligent Computer Technologies
Department
4
Course Outline

• Part 1. Introduction to Data and Knowledge
  Engineering (12 hours, 8 tests)
• 1.1. Working with information data and knowledge
• 1. 2. Knowledge representation models
• 1. 3. Working with fuzzy information
• 1. 4. Knowledge engineering structure
• 1. 5. Psychological aspect of data and knowledge
  acquisition
• 1. 6. Linguistic aspect of data and knowledge
  acquisition
• 1. 7. Methodological aspect of data and knowledge
  acquisition

• Part 2. Practical Knowledge Engineering (16
  hours, 10 tests)
• 2.1. Individual communicative data and knowledge
  acquisition methods
• 2.2. Group communicative acquisition methods
• 2.3. Textological methods
• 2.4. Data and knowledge structuring
  object-structured analysis
• 2.5. Ontologies and visual modelling
• 2.6. Functional distribution in systems project
  team

5
There are forms and rythms that are hidden for
the eye of a contemplator, but opened for the eye
of an analyst.R.Feinman
6
Tatiana Gavrilova
????-?????-???????
1. Introduction to Intelligent System Development
2. Knowledge Engineering 3. Knowledge Management
7
1. Introduction to Intelligent System
Development1.1. Knowledge Engineer
8
Knowledge Engineer
Communicative Abilities
Analytical Abilities
9
(No Transcript)
10
Who is Analyst?

• Problem Originator
• Knowledge engineer
• Cognitive Engineer
• Chief Information Officer (CIO)
• Chief Knowledge Officer (CKO)
• Knowledge Manager
• Knowledge Broker

11
1.2 Short History of AI

• AI - as a basement of NIT
• AI Definition
• 2 branches and main directions

12
New Information Technology vs Traditional
Information Technology
13
??????? ?????????
????-?????-???????
First steps in AI research

• 1956 - USA Dartsmouth's College, Massachusetts
  Artificial intelligence
• (not intellect)
• 1954 - MGU, workshop Automates and mind

14
Artificial Intelligence
15
Short history of AI

• Neuro-cybernetics
• 50-70 - neural networks, PERCEPTRON
• 70-90 - scepticism, Lighthills commission
• 90-2000 - neuro-computers, industrial software

• Black box
• 50-60 - maze scheme
• 60-70 logic approach
• 70-90 expert systems
• From 90- industrial AI

16
Situation and Prospect of AI Market
17
AI Market
18
Situation and Prospect of AI Market
19
Situation and Prospect of AI Market
20
Main IS Applications
21
AIs main research fields
22
Test N1(trains)
23
1.3. Data and Knowledge

• Data and Knowledge
• Concepts Intensional and Extensional
• Knowledge Classification

24
Information
DATA - instances and facts characterizing
objects, processes and their properties
KNOWLEDGE - rules, which are concatenating the
data and that are obtained from experience.
25
DATA
Data are instances, characterising objects,
processes and phenomenon of the subject domain as
well as their properties
During the computer processing the data is
transformed in the next main stages
D1 Data as the result of observing and measuring
D2 Data on the material objects (tables,
records, references)
D3 Data Models (Structures) in the form of
diagrams, plots, functions
D4 Data in computer on the data description
language
D5 Electronic digital information store
26
Knowledge
Knowledge are the objective laws (foundations,
rules, links) of the subject domain, obtained as
the result of the practice, professional
experience and allowing the specialists to set
and solve problems in this domain. Knowledge is
well-structured data or data about data or
meta-data.
27
Knowledge
During the computer processing the knowledge as
the data is transformed in the next main stages
Z1 Knowledge in the persons memory as a result
of the reasoning
Z2 Material object (manuals, training aids)
Z3 Knowledge field conditional specification
of the main subject domains objects, their
attributes and regularities between them
Z4 Knowledge, defined at knowledge
representation languages (production languages,
semantic networks, frames)
Z5 Knowledge Base at computer information
carriers.
28
Knowledge classification

• shallow knowledge
• deep knowledge

29
Natural automation
All management solutions are based not on data,
but on knowledge
30
Intensional vs. Extensional

• Intensional definition of a concept is knowledge
  based on high level of abstraction ( ex. a car
  could be said to be a vehicle of transportation
  that has four wheels and an engine) .

• On the contrary extensional definition is based
  on data and specific examples (ex. car Ford,
  Opel, Benz etc.)

31
Test N2(extensional /intensional)
32
1.4. Knowledge Representation Models
33
Classification of Knowledge Representation Models
34
Semantic Nets
Semantic net is an oriented graph with nodes are
representing objects/concepts and arrows are
representing relations between them.
If the relation connects two nodes the net is
called bi-net and in case of multiple
relationship, n-net.
35
Semantic Network
value
Red
Color
Engine
has
feature
part
is
is
Mode of transport
Volga
Automobile
like
belong to
for example
Ivanov
Person
36
Types of relationships

• Hierarchical (A-Kind-Of, Is-A)
• Causal (if- then)
• Quantitative (more than, equal)
• Functional (runs, eats, is)
• Spatial (on, behind, inside)
• Temporal (after, before, until)
• Attribute (colour)
• Value (red)
• Structural (has-part)

37
Test N3(Semantic Network)
38
Frames
The concept of FRAMES was proposed by Marvin
Minsky in 1972.
.
Frames are used as an abstract structure for
representation of stereotypes of complex
objects, processes, events, and scenarios.
Frames could also be used to model rules and
stereotypes in behaviour.
39
Frames Name ROOM
Slot-AKO (?-kind-of) ACCOMODATION, Slot 2, area
7-30 sq.m , Slot 3, height 2.5 - 4 m., Slot 4,
floor , Slot 5, ceiling , Slot 6, window
, Slot 7, furniture Slot 8, walls color
40
Structure of the Frame
41
Frame Network
42
Test N4(Frames)
43
Rule-based model

• Rule is a sentence which consists of two parts,
  the first being the condition and other being a
  result or an action.
• IF condition, THEN action

44
Production Model

• R1 IF business trip is sanctioned ? head of
  department agrees, THEN pay advance
• R2 IF there is contract or expectancies
  record, THEN business trip is allowed

45
Interpreter Working Cycle
Rules Selection criterion
Comparison
Conflict set
Conflict resolution
Working memory (database)
Rule database
Action
Invoked rule
46
Interpreters functioning scheme
47
Deduction strategies backward deduction
Depth Search
48
Deduction strategies direct deduction
49
Test N5(production model)