Adaptive User Interfaces
Motivation
Project
Want to help?
References
Particular Project
(See also Poster
Slides for CasCON
)
-
Prediction of User Command (unix)
-
latex foo.tex
dvips foo -o
ghostview foo.ps
-
gcc bar.c
link -o blah ... bar.o
blah
-
Steps:
-
Build model of user's command sequences
by watching sequences of earlier user commands
-
Exploit this model
-
Load predicted command into buffer
-
Set ``F1'' to most likely, ``F2'' to 2nd, ...
-
Create new macro with sequence
-
For more information, read
Predicting UNIX
Command Lines: Adjusting to User Patterns
How you can help?
To try out our ideas, we need DATA:
we need to know what sequence of commands users -- real
users -- really type.
If you are willing to help us collect data, follow these instructions
in this webpage.
(If you have some other type of Unix box [linux, solaris, ...], contact
benjamin@cs.ualberta.ca)
Notes:
-
You can decide when to let your sessions be recorded, and can turn
it off anytime you wish.
-
If you request: we will make your sessions "anonymous".
-
Once we collect the data, we can make it available to the participants.
(After "sanitizing it", as appropriate.)
Motivation
-
EVERYBODY is using computers
-
Most applications are interactive:
-
editors
-
spreadsheets
-
web-browser/searchers
-
games
-
DBMS
-
expert systems
-
...
=> It is important that computers be easy to
use
-
As different users are different
different background knowledge, styles,
preferences, ...
one approach is NOT sufficient.
-
But most software systems treat all users as the same!
"One size fits all"
-
Yes, many programs can be customized
... by setting some high-level parameters, ...
But this process is
-
Manual
-
Tedious
-
At wrong ``level''
-
Requires user to be aware of preferences
-
Better: Build systems that can adapt to user
-
Goal: Software system that itself
to user
by WATCHING user's behaviour
Efficiently and easily (innocuously)
acquire
accurate model of user's interests/goals/...
and then use it to simplify interactions
Additional Information
If you want to see other related research on this topic, see
and/or the other material listed in 28/Aug AI-Seminar
(eg, the presentation
(4ps)).
See also the journal
Or contact
R Greiner
greiner-at-cs-dot-ualberta-dot-ca
Benjamin
Korvemaker benjamin@cs.ualberta.ca