Presentation Outline Motivation Basic concept Bakground Futur work Conclusion Nesrine MEZHOUDI nesrine.mezhoudi@uclouvain.be User Interface Adaptation.
Post on 26-Dec-2015
213 Views
Preview:
Transcript
1
Presentation Outline
Motivation
Basic concept
Bakground
Futur work
Conclusion
Nesrine MEZHOUDI nesrine.mezhoudi@uclouvain.be
User Interface Adaptation Based on User Feedbacks and Machine
Learning
Louvain Interaction LabUniversité catholique de Louvain
Promotor:Prof. Jean Vanderdonckt
Jean.vanderdonckt@uclouvain.be
2
AdaptationUser-centered adaptation
3
Adaptation User-centeredness
4
Adaptation User-centeredness
5
Outline
Motivations
Basic concepts
Methods & Application
6
Problem: adaptation rules are static
Adaptation rules are implemented according to a predefined and static set of standards, guidelines, and recommendations
Hardly re-adaptableBarely impossible to updateHighly expensive (redevelopment, time, human resources)
7
Problem: static rules prevent adaptation
• Dissatisfaction• Frustration• Discouragement• Loss of motivation• …
8
Solution: involve the end-user in the user interface development
Throughout the system life-cycle From the early stages of the system life-cycleStarting from the user interface definition
9
10
Well-rounded feedback topology
Implicit Feedback
Explicit Feedback
Without rating aims With rating aims
10
11
Unified theoretical architecture for adaptation based on ML
Context• User• Platform• Environment
Adaptation Rules
Repository
Adaptation Management
Layer
Perce
ptio
n(tra
cking
tools, se
nso
rs…)
Recommendation
FeedbackR
ein
force
men
t
Evalu
atio
n
Updatin
g Adaptin
g
Perc
eptio
n La
yer
UI
12
Adaptation Rule Manager
Adaptation Rules
Repository
Trainer-Rule Engine
Learner-Rule Engine
Generated Rules
Rule Engine
Rule Management
Tools
Training Rules
Feedbacks
User
13
Adaptation Rule Manager
Adaptation Rules
Repository
Trainer-Rule Engine
Learner-Rule Engine
Generated Rules
Rule Engine
Rule Management
Tools
Training Rules
Feedbacks
User
(1) Executing pre-existed adaptation rules, serving as a training set to (2) detect a pattern of user behavior throughout his feedbacks. Besides, (3) coming up with statistics and (promote/demote) ranking for the Learner Rule Engine (RLE).
14
Adaptation Rule Manager
Adaptation Rules
Repository
Trainer-Rule Engine
Learner-Rule Engine
Generated Rules
Rule Engine
Rule Management
Tools
Training Rules
Feedbacks
User
analyzing collected user judgments. Which are intended to serve in a promoting/demoting ranking, Then generate new decision rules , (Learns)
Potential applications
15
Tasks
AUI
CUI
Final UI
Potential applications
16
Tasks
AUI
CUI
Final UI
17
Time-line
State of the arts
Conceptualization
Implantation
Test & Evaluation
18
Thank you for your attention
Nesrine Mezhoudi nesrine.mezhoudi@uclouvain.be
top related