KI-CBR + BNs

Download Report

Transcript KI-CBR + BNs

Integrating case-specific experiences
with general domain knowledge
for intelligent information processing.
Agnar Aamodt
Norwegian University of Science and Technology (NTNU)
Department of Computer and Information Science (IDI)
Division of Intelligent Systems (DIS)
Artificial Intelligence and Learning Group (AIL)
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
The Issues
•
Case-based reasoning (CBR) intro
- basic, and knowledge-intensive
•
Knowledge modeling and representation
- Merging general and situation-specific knowledge
•
Reasoning: Knowledge-intensive case retrieval
- Syntactic and explained matching
•
Development and testing tools
- Knowledge modeling editor and case matching visualizer
•
Recent and ongoing research
•
Challenges ahead
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Case-Based Reasoning
•
CBR is a technology for solving a new problem by retrieving a similar case whose
solution is known, and adapting its solution.
Past Solution
New Solution
Past Problem
New Problem
Case-1
General knowledge
•
It is also a technology for machine learning, by integrating a new case just solved into
the exisiting case base.
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Knowledge-Based Methods
- Development History
Heuristic
rules
Rule-based systems
(e.g.: MYCIN)
-> Rule-based reasoning
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Knowledge-Based Methods
- Development History
Control Knowledge
Heuristic
rules
Explicit Control Knowledge
(e.g. NEOMYCIN)
-> Meta-level reasoning
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Knowledge-Based Methods
- Development History
Control Knowledge
Heuristic
rules
Deep knowledge
Deeper models, text book knowledge
(e.g.: CASNET)
-> Model-based reasoning
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Knowledge-Based Methods
- Development History
Control Knowledge
Specific
cases
Heuristic
rules
Deep knowledge
From generalized to situation-specific knowledge (e.g. PROTOS, CHEF)
-> Case-based reasoning
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Knowledge-Based Methods
- Development History
Control Knowledge
Specific
cases
Heuristic
rules
Deep knowledge
Integrated systemes
(e.g. SOAR, THEO, META-AQUA, CREEK)
-> Architectures for intelligence
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
The Case-Based Reasoning Cycle
Problem
New
Case
RETAIN
Learned
Case
Retrieved
New
Case
Case
Past
Cases
General
Knowledge
Tested/
Repaired
Case
Solved
Case
Confirmed
Solution
Suggested
Solution
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
The ”knowledge-intensiveness” scale of CBR
• No explicit gen. knowledge
• A lot of cases
• A case is a data record
• Substantial gen. knowledge
• Not very many cases
• A case is a user experience
• Simple case structures
• Global similarity metric
• No adaptation
• Learning is simple storage
• Complex case structures
• Sim. assessm. is an explanation
• Knowledge-based aptation
• Knowledge-based learning
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
What is this thing called Knowledge?
Focus: A decision-making process:
Decision
Step
Input
• reasoning agent
Output
ENVIRONMENT
• interactive world
• reasoning agents
2 important perspectives on knowledge:
- its role
- its frame of reference
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Data - Information - Knowledge
Knowledge
Interpreted sym bol structures
- use d to interpret data , e laborate on information, a nd learn
- use dwithin decision steps
Learning
Elaboration
Information
Interpreted sym bols a nd sym bol structure s
- input to a dec ision ste p
- output froma decision step
Data
Interpretation
Data
Obse rved, uninterpreted sym bols
- signs, c harac ter sequenc es, patterns
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Knowledge and cases
So, why do CBR systems need general domain knowledge
to become knowledge-intensive?
Aren’t cases knowledge themselves?
1. Cases are sometimes viewed as data/information, not knowledge.
2. Cases are often viewed as shallow knowledge. Additional general
knowledge deepens - and widens - the system’s knowledge.
3. Cases are knowledge, but their knowledge content is enhanced by
their features being defined and related within a body of general
knowledge.
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
The Creek Approach
1
•
Combines case-based and model-based reasoning,
for problems in open and weak theory domains.
•
Input is problem solving context (e.g. goal) and
problem features (e.g. a list of findings).
Output is the best plausible interpretation of the input
within the context.
•
Knowledge types, used for reasoning are
-a body of situation-specific knowledge,
i.e. a case memory of findings linked to
solutions, annotated with other relevant
information and knowledge
-a body of general domain knowledge, as deep
relationships or heuristic rules
1
Case-based Reasoning through Extensive Explicit Knowledge
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Model-Based Reasoning
MMo
•
MBR - in this context - is a technology for solving a new problem by explaining its
solution within a multi-relational model of the target system.
thing
goal
hsc
case
hsc
domain-object
hsc
hsc
hsc
diagnosis
find-treatment
find-fault
hsc
has-output
described-in
vehicle
case#54
has-function
hi
van
transportation
hsc
hsc
has-status solved
diagnostic-case
tested-by
car
hp
hd
hp
wheel
test-procedure
possible-status-of
hp
test-step has-electrical-status
hp
hp
engine
hi
has-state
hsc
starter-motor-turns
has-engine-status
has-fault
tested-by
hsc
fuel-system
case-of
electrical
diagnostic-hypothesis
-system
engine-test hsc
N-DD-234567
has-fault
engine-turns
hp
car-fault
hsc
test-for
battery-low
subclass-of
has-fault engine-fault
fuel-system-fault
hsc
instance-of
hsc
battery
hsc broken-carburettor-membrane
subclass-of
subclass-of
has-fault
electrical-fault hsc
status-of
part-of
battery-fault
observed-finding
tested-by
hsc - has subclass
subclass-offinding
turning-of
hi - has-instance
test-for
hp - has-part
-ignition-key
starter-motor
hd - has-descriptor
•
hsc
MBR - in this context - involves the abductive steps of hypothesis generation and
evaluation/selection, for which methods of plausible reasoning are applied.
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
thing
goal
hsc
hsc
hsc
find-treatment
find-fault
has-function
transportation
hsc
car
hsc
diagnosis
hsc
has-output
described-in
vehicle
hi
van
hsc
diagnostic-case
tested-by
has-status
test-procedure
hp
has-fault
fuel-system hp
has-fault
electrical
hsc
tested-by
ca rburehsc
ttor
has-fault
hs c
fuel-s
ys tem-fa ult
has-fault
hs c
hsc
hp
causes
hi
starter-motor-turns
condens ation- in-gas -ta nk
engine-turns
test-for
engine-fault
causes
battery-low
subclass-of obse rve d-f inding
wate r-in-ga s-mixt ure
instance-of
battery
causes
has-fault
no-chambe
r-ignition
hi
subclass-of
subclass-of
hi
causes
status-ofpart-of
battery-fault
N-DD-234567
hs c
too-r ich-ga s-mixt ure -in- cy linder
enigne-t urns
engine-doe s-not- fire
observed-finding
tested-by
- has subclass
- has-instance
- has-part
- has-descriptor
case-of
obse rva ble -sthsc
ate
engine-test
wate r-in-ga s-t ank
has-fault
ca rburehsc
ttor -va lve -st uck
hsc
causes
hsc causes
broken-carburettor-membrane
causes
fuel-system-faulths c
hsc
has-state
has-engine-status
hi
car-fault
ca rbure ttor -va lve -fa ult
electrical-fault
has-electrical-status
diagnostic-hypothesis
ca rbure ttor -fa ult
hs c
solved
possible-status-of
test-step
hp
engine
fuel-s ys tem
hsc
hi
hp
hd
case#54
hd
hp
-system
case
hsc
domain-object
hp
hp
wheel
hsc
test-for
starter-motor
turning-of
subclass-of finding
-ignition-key
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
The Creek Architecture: Basic Knowledge Types
thing
generic concepts
general
domain concepts
cases
case case case
039 76 112
1
Case-based Reasoning through Extensive Explicit Knowledge
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Knowledge engineering as modeling
Task
Reality
Modeling
(Sub) Pr oblem Descri ption
Model
of
Task
Reality
(Parti al) Solution to Pr oblem
Task reality
Agent:
: The entire spatial and temporal extension of the world
which is relevant for accomplishing a real-world task.
The physical entity that, embedded in a task reality,
accomplishes a task.
Task :
What is to be accomplished
(goal, purpose).
Method
: How a task is accomplished
(procedure, control).
Dom ain knowledge
:
Possessed by agents and used within methods
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Example:
Knowledge-level modeling -> The Components of Expertise framework
domain
model-1
domain
model-2
user
case
model-1
Task Decomposition
model construction
activity
case
model-2
Model Dependency Diag ram
Methods:
• Decompose tasks
• Execute tasks
• Assign tasks to model
construction ativities
• Impose control over tasks
T ask-2
T ask-1
c2
T ask-3
c1
Control Diag ram
Knowledge engineering as modeling
Problem solving as modeling
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
The kn owledge modeling cycle
Knowledge Level
Analysis and Modeli ng
ME NTA L
MODE L
CONCEPTUA L
KNOWLEDGE
MODE L
Symbol Level Design
and Implementation
COMP UTE R
INTERNAL
MODE L
Ini tial K now ledg e
Mo delin g
Kn owle dge
Ma inten anc e
Problem
Solving
New Case
EX PE RIE NCE
Periodi c
Knowledge
Revision
Sustained
Learning
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Creek Top Level Ontology
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
The Creek Explanation Engine
• Goal
- Appl. task is defined
• Situation
- Findings are listed
- Constraints are specified
- Solution asked for
• Goal
- Appl. task accomplished
• Situation
- Findings explained
- Constraints confirmed
- Solution found
FOCUS
ACTIVATE
EXPLAIN
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
The explanation engine within the CBR cycle
Retrieve
Activate
Explain
Reuse
Focus
Activate
Explain
Revise
Focus
Activate
Explain
Retain
Focus
Activate
Focus
Explain
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Index structure
relevance factor
CASE 1
FINDING A
CASE 2
FINDING B
CASE 3
FINDING C
CASE 4
SOLUTION A
CASE 5
SOLUTION B
CASE 6
CASE 7
General domain k nowledge
- inf erring of f indings
- inf erring of s olution c lass
•
Cases are indexed by
- relevant findings
- differential findings
- solutions
•
Numerical features are indexed with their
least general class or range.
•
No explicit index hierarchies
•
Implicit abstract indexing through possible
inferring of abstract findings
(observed and inferred)
(only one different finding)
(faults and treatments)
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Case indexing
no-lights
subclass-of
relev ant -f eature-in
case-19
has-relev ant-f inding
Predictivestrength
sufficient
strongly-indicat ive
strongly-indicat ive
strongly-indicat ive
indicative
...
spurious
elsystem-f inding,v isual-observ at ion
(case-19 0.8) (case-213 0.95) (case-466 0.85) (case-351 0.9)
(no-lights
:import ance necessary
:predictiv e-strength indicat iv e)
Importance
<any>
necessary
charact eristic
informative
necessary
...
irrelevant
Relevancefactor
1.0
0.95
0.90
0.85
0.80
...
0.0
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Similarity - Phase 1, direct matching
n
m
  sim( f , f
i
sim(CIN ,CRE ) 
i1 j 1
j
) *relevance factorf j
m
 relevance factor
fj
i1
C IN and CRE are t he input and r etri eved cases , n i s the number of findings in C IN, m is the number of
findings in CRE , f i is the ith finding i n C IN, fj the jth finding i n C RE , and s im(f1,f2) i s s imply given as:

1
sim( f1 , f 2 )  
0
if f1  f2
otherwise
The relevan ce f actor is a nu mber t hat combines the predictive s trength (d egree of sufficiency)and
importance (deg ree ofne ces sity) of a feature for a stored cas e.

ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Similarity - Phase 2, indirect (explained) matching
The paths have convergence points, i.e. explanatory concpets - s uch as causal concepts, for which there
exist an explanation path from both findings. Its strength is the product of the strength of each relation
leading from the finding to the convergence point:
n
path strength ( f , c)   relation strength i
i 1
Here, n is the number of relations. There may exist one or more parallel paths from each finding to each
convergence point. The resulting strength is based on the general formula for adding contributions fr om n
parallel elements, S1... Sn, into a total score:
n
parallel strength(S1,S 2 ,..S n )  1  (1 Si )
i1
Thus, the total combined strength of all the paths leading from a finding f to a convergence point c, with n
being the number of paths between f and c, is comp uted according to the following formu la:

n
total path strength( f ,c)  1  (1 path strength( f ,c) i )
i1
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Similarity - Phase 2, indirect (explained) matching
The st reng th of one exp lanation path ( eps) leading fro m a f inding f1 to a finding f 2 via t he conv ergen ce
point c, is computed by multiplying the t otal path s treng th for each of the findings to t he convergence point,
and t heto tal exp lanation s trength for the two findings (f1 and f2) vi a se veral converg ence points is finally
computed byu s ing the parallel s trength for mula:
eps( f1, f2,c)  total path strength( f1,c)  total path strength( f2,c)
n
explanation strength( f1 , f 2 )  1  (1 eps( f1 , f 2 ,c) i )

i1
Here n is the number of converg ence points between the findings, and c i is the ith convergence poin t.

ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Main network inference approach:
Plausibel Inheritance
(Cohen and Loiselle, 1988)
Car
has component
Engine
subclass of
Alcohol
Intoxication
causes
has component
contains
Sportscar
causes
Coughsyrup
standard method
extended method
has part
car
requires
has part
engine
piston
oil
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Example of Plausible Inheritance
associated
with
bacterial
epidemic
used for
drinking
subclass
of
subclass
of
caused by
caused by
epidemic case #3
water
bacterial
infection
Inheritance rules:
I = ((subclass of, causes),
(subclass of, caused by),
(subclass of, associated with),
(subclass of, used for),
(subclass of, has solution),
(causes, causes),
(caused by, caused by),
(caused by, has solution)
(associated with, associated with))
has solution
dirty water
associated
with
clean water supply
bad hygiene
Formalism:
Tt set of relation-types transferable
through relation t
(ct) set of relation-types transferable
through concept c along t
Ri set of relationships inherited to
the initial concept
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Example of Plausible Inheritance
Initialization
associated
with
bacterial
epidemic
used for
drinking
subclass
of
subclass
of
caused by
caused by
 = {t}
epidemic case #3
water
bacterial
infection
has solution
dirty water
clean water supply
associated
with
bad hygiene
Question: Find the extended frame for
’epidemic case #3’
or
Check if ’clean-water-supply’
can be a solution to ’epidemic case #3’
or
Find plausible solutions
for ’epidemic case #3’
Ri = {}
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
First step
Tsubclass of ={
causes,
caused by,
used for,
has solution,
associated with }
bacterial
epidemic
Tcaused by ={
subclass
of
caused by,
has solution }
caused by
epidemic case #3
I = ((subclass of, causes),
(subclass of, caused by),
(subclass of, associated with),
(subclass of, used for),
(subclass of, has solution),
(causes, causes),
(caused by, caused by),
(caused by, has solution)
(associated with, associated with))
bacterial
infection
={
causes,
caused by,
used for,
has solution,
associated with }
={
caused by,
has solution }
Ri =
{(‘epidemic case#3’, subclass of, ‘bacterial epidemic’),
(‘epidemic case#3’, caused by, ‘bacterial infection’)}
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
(ct) = Tt I (ct-1)
={
causes,
caused by,
used for,
has solution,
associated with }
={
associated with }
associated
with
bacterial
epidemic
caused by
caused by
epidemic case
#3
used for
drinking
subclass
of
subclass
of
=T
water
bacterial
infection
I = ((subclass of, causes),
(subclass of, caused by),
(subclass of, associated with),
(subclass of, used for),
(subclass of, has solution),
(causes, causes),
(caused by, caused by),
(caused by, has solution)
(associated with, associated with))
has solution
dirty water
associated
with
={
caused by,
has solution }
clean water
supply
bad hygiene
Ri =
{(‘epidemic case#3’, subclass of, ‘bacterial epidemic’),
(‘epidemic case#3’, caused by, ‘bacterial infection’),
(‘bacterial epidemic’, associated with, ‘dirty water’)}
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
={
causes,
caused by,
used for,
has solution,
associated with }
={
caused by,
has solution,
associated with }
bacterial
epidemic
subclass
of
associated
with
epidemic case
#3
bacterial
infection
={
caused by,
has solution }
I = ((subclass of, causes),
(subclass of, caused by),
(subclass of, associated with),
(subclass of, used for),
(subclass of, has solution),
(causes, causes),
(caused by, caused by),
(caused by, has solution)
(associated with, associated with))
water
used for
subclass
of
has solution
caused by
dirty water
caused by
=T
={
caused by,
has solution }
associated
with
drinking
=
clean water
supply
bad hygiene
={
associated with }
Ri =
{(‘epidemic case#3’, subclass of, ‘bacterial epidemic’),
(‘epidemic case#3 ’, caused by, ‘bacterial infection’),
(‘bacterial epidemic’, associated with, ‘dirty water’),
(‘bacterial infection’, caused by, ‘dirty water’),
(‘dirty water’, has solution, ‘clean water supply’),
(‘dirty water’, associated with, ‘bad hygiene’)}
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Creek Top Level Ontology
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
CREEK
•
Retrieve
- context focusing by spreading activation in
the semantic network, followed by
- index retrieval of possible cases, followed by
- explanation-driven selection of best match
•
Reuse
- attempts to copy solution from matched case
- explanation-driven adaptation, by combining
explanantion of retrieved case with general
domain model
•
Revise
- user evaluates and gives feedback
- case status info kept and used in case
selection and reuse
•
Retain
- attempts to merge the two cases
- stores relevant findings, successful and failed
solutions, and their explanations
- updating the strength of indexes
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
•
Ongoing projects (2003– Knowledge-based learning support (IKT & Læring, NTNU)
– AmbieSense – Context-sensitive info for mobile users (EU)
– COST 282 Action – Knowledge exploration in sci & tech (EU)
– Explanation of gene-gene relationships (Bioinformatikk, NTNU)
– Avoiding unwanted events in drilling (Internal)
– AI systems in the working environment (Internal)
– From text to structured cases - data/text mining (Internal)
– Conversational CBR for software component retrieval (Internal)
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Avoiding unwanted events in drilling
With Pål Skalle, Martha Dørun Jære, Inge Valaas, Elise Bakke, Tore Brede
The Goal
Predicting possible unwanted events during a drilling operation, in order to steer
away from them and hence avoid them.
Overview
Temporal case-based reasoning - based on temporal intervals - is used to
represent cases. The CBR process supervises the drilling prameters, and gives a
warning whenever a situation occurs that reminds of a past situation that failed.
If so, CBR and MBR are used to suggest how to avoid the potential unwanted
event.
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Schematic view of a North Sea Well
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
From data to temporal intervals
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Matching of temporal cases
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Component Retrieval Using Conversational Case-Based Reasoning
With Mingyang Gu and Torgeir Dingsøyr
The Goal
A method that will combine knowledge-intensive CBR with the Conversational
CBR method, applied to software component retrieval from component libraries.
Overview
Component retrieval, about how to locate and identify appropriate components,
is one of the major problems in the component reuse. And it becomes more
critical as more reusable components come from component markets instead of
from an in-house component library, and the number of available components is
dramatically increasing. In CCRM, components are represented as cases, a
knowledge-intensive case-based reasoning (CBR) method is adopted to explore
semantic similarity between users’ query and stored components, and
conversational case-based reasoning (CCBR) technology is selected to acquire
users’ requirements interactively and incrementally
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
The object-level knowledge model
Component query (partial)
has-number-of-parameter1
has-image-color-apaceRGB
has-image-dimension3D
has-image-file-type BMP file
has-errorfile-open-error
has-file-size-constraints5 megabyte
…
Write BMP component (partial)
has-number-of-parameter 1
has-image-color-spaceXYZ
has-image-dimension2D
has-image-file-type BMP file
has-errorfile-open-error
has-file-size-constraints(and (> 0
Bytes)
(< 100 Megabyte))
…
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
The meta-level knowledge model
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Bayesian text mining for case generation
With Stig Hollund, Helge Langseth, and Rob Engels
The Goal
A method that will take a text report of an episode (e.g. a part of an oil well
drilling report or a medical patient journal), and construct a case description
in the Creek representation formalism.
Overview
CBR systems typically stores and reasons from cases that have a well-defined
structure, i.e. a case contains certain types of features that take certain types of
values. Textual CBR systems, on the other hand, stores cases as plain text, and
uses text-based information retrieval methods for retrieving them. However,
such cases are hard to interpret in terms of content, and hence hard to use for
adaptation, targeted retain methods, etc.
Our method is different: It aims to construct a Creek type case structure from
plain text, using Bayesian methods combined with knowledge of a case’s
structure. The text mining method will be compared and related to CognIt’s
Corporum Summarizer.
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Knowledge construction and maintenance
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
With Wacek Kusnierczyk, Astrid Lægreid, and Arne Sandvik
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Concept Map Matching in Tutoring
With Frode Sørmo, Arvid Holme, Pål Skalle
Hypothesis:
Similar Concept Maps predict similar competence in solving
exercises.
.
Experiment:
About 75 1st year student in Java course model their theoretical
knowledge and solve exercises. CBR used to match Concept
Maps and attempt to predict exercise difficulty.
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Students model their own theoretical
knowledge in Concept Maps
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
CREEK-ITS (Frode Sørmo)
New Case
Jane
Predicted Exercise
Competence
Peter
Mary
George
Peter
Concept Map Space
George
Jane
Exercise Competence Space
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Intelligent Systems in the Working Environment
With Jörg Cassens, Wolter Pieters, Peter Svedberg
Goal:
To investigate how intelligent systems behave different from other
computer systems in working environments, and what consequences
this has for the development of AI systems.
.
Overview:
Study the use of the following three perspectives within this context:
- Actor Network Theory
- Activity Theory
- Semiotics
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
The Action Cycle
Action Cy cle
Task
Initial goal for
a task is set
Goal Setting
Feedback
Controlling whether
goal was achieved;
if not, why?
Decide how to achieve the goal
Includes the selection of tools
and actions
Planning
Consciously
performed
Action
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
AmbieSense – Ambient and context-sensitive info to mobile users
Small and wireless context tags
- Holds contextual information about the current situation around
it
- Can be remotely administrated by
authorisedusers
- Context information can be changed
– also the earth-position
- Works both independently and together with wireless networkndards
sta
- Automatically helps the user to define the current context;
=> Achieves usability and user
-friendliness without additional effort for the
mobile users
Business lounge
Restaurant
Statue
Shopping center
Information service providers providing
information to the right situation (context)
- Tourist info for a city center area
- Travellersat an international airport
- Personalised
, mobile infotainment
- Restaurant, menus, hotels guide
- Shopping and advertisement
=> Achieves more differentiated and
personalised
distribution of information
Mobile/ travelling users that get
right information in the right situation
- Tourists in a city center– shopping, restaurants etc)
- Travellersat the international airport
- Travellersthat need personalisedinfotainment
- ..
The AmbieSense mobile computer
- Wireless network access both indoor and outdoor (IEEE, GSM,S,GPR
etc)
- Context tags technology automatically helps user to identify
e current
th
user context
- User context software administrates the various user contexts
nd a
the information units linked to
them
- Intelligent and mobile agents also helps to automatically defin
e the current context
- The users can manually specify the context by themselves;
=> achieves usability, customer satisfaction, more differentiated
nd personalised
a
information space
= Context tag mounted anywhere in order to help the mobile userget right information in the right situation
= Wireless communication
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
The AmbieSense Consortium
• SINTEF Telecom and Informatics, N (Project Coordination, CO)
• Robert Gordon University (Academic Partner, CR)
• Norwegian University of Science and Technology (Acad. Partner, AC)
• Siemens AG (Industrial Partner, Technology Provider, CR)
• CognIT a.s (Industrial Partner, Technology Provider, CR)
• Reuters (Industrial Partner, Test Cases and Information bases, CR)
• YellowMap (Industrial Partner, Information and Service Provider, CR)
• Oslo Airport, (Case-study partner, AC)
• Seville Global (Tourism, Case-study, AC)
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Summary
• Knowledge is a resource/asset for reasoning,
with particular roles and a particular frame of reference
• Knowledge-intensive CBR enhances the knowledge content
of a case
• Knowledge-level modeling helps build operational
knowledge models
• Visualization and interactive tools help build systems
and gain confidence
• The Creek architecture, tools, and reasoning systems help understand
core issues related to knowledge-intensive CBR
However:
- An experimental evaluation environment is missing
- A deployed application is missing
--> Efforts under way
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt
Thanks
for your attention
ICIIP 2004, Beijing, Oct.21-23, 2004. Agnar Aamodt