KIPA Lectures Summary

KIPA summary
Lecture 1: Introduction
Knowledge systems

- A Knowledge System is a system that uses knowledge to perform a task
- This knowledge is explicit (communicated as is)
o In contrast to machine learning, knowledge there is implicit (data)

Objectives of knowledge systems

- Knowledge capitalization
o Learning from the past by knowledge retention and re-use
- Project accompaniment
o Learning from present activities by knowledge sharing and exchange
- Innovation
o Moving towards future benefits by leveraging organizational knowledge assets
- Cost reductions
o Achieving cost reductions through first-time right adoption enabled by knowledge
sharing

MYCIN

- Inference engine
o Processing information in KB
o Using information to make predictions and recommendations
o Using information in the KB to determine what questions to ask
- Knowledge base
o Information that MYCIN uses to make decisions
o Does not control the flow of the system

Task types

- Two main task types
o Analytic: use knowledge to inform a user
o Synthetic: use knowledge to create a product or service

Lecture 2: Knowledge Modelling
What is knowledge

- Knowledge is ability or directions to interpret and use information for specific tasks
o Knowledge is complex information
o Rules and general facts

Knowledge vs Information

- Information
o Specific facts
- Knowledge
o General facts and rules

Knowledge base

, - Domain knowledge is stored in a knowledge base
- A knowledge base contains
o Concepts
o Attributes
o Relations
o Rules
- The KB is based on a domain schema

Rules and inference

- Knowledge can instantiate rules
o Analytic task: rules to derive conclusions
o Synthetic task: rules to determine which actions to perform

Rule structure

- A rule models a relation between values of properties
- Structure: property+value RELATION property+value

Problem solving

- Make a decision or a case based on norms

Example analytic:

- decide on a loan for a person
o Case (loan application)
o Decision category (elibigle or not)
o Norms (rules for income)
- Input: case, case-specific norms
- Output: decision

Example synthetic:

- Configure a computer system
o Component (hard disk)
o Parameter (disk capacity)
o Constraint (graphics card X supports only one screen)
o Preference (include GPUs)
o Requirement (price ≤ ¤1000)
- Input: set of requirements
- Output: set of components and parameter values

Lecture 3: Knowledge Acquisition
Knowledge acquisition

Knowledge acquisition steps

- Knowledge identification
o Domain familiarization
o Identify task- and domain-related components
- Knowledge specification
o Task description

, o Domain concepts
o Knowledge Base construction
- Knowledge refinement
o Validate model with simulation or prototype
o Modify knowledge base when needed

Information sources

- Nature of information sources
o Theoretical, well-developed, diffused, . . .
- Diversity of information sources
o Different experts and other sources
o Conflicting sources
o Practical constraints on using sources

Specification

Task templates

- Choose task template based on various aspects
o Output: category, decision, plan
o Input: type of task data and domain data
o System: process, machine, living entity
o Presence of constraints and requirements
- SIMPLE TASKS DO NOT HAVE TASK TEMPLATEs
- COMPLEX TASKS DO HAVE TASK TEMPLATES

Domain schema

- Domain-specific concepts
o Independent of the task
o Not likely to change
- Task-specific elements
o Problem solving rules
o Task concepts




Specification procedure