A Bangla Predictive Keyboard For People With Neuro-Motor Disorders

1
A Bangla Predictive Keyboard For People With
Neuro-Motor Disorders

• Presented By
• Animesh Mukherjee
• Research Scholar
• Department of Computer Science and Engineering
• IIT Kharagpur

2
The Reality

• Suppose you are asked to use a computer which has

3
The Divide

• Nevertheless there is a big population in India
  (14.56 million approx) that experiences such a
  difficulty every day
• These are people suffering from neuro-motor
  disorders
• For them the presence or the absence of a mouse
  or a keyboard is always synonymous to its absence

4
Neuro-Motor Disorder What is it ?
Produces Nerve Cells that Causes Movements of the
Body Parts
Serves to Modify the Movements

• These disorders are caused by -
• Faulty development of motor areas in the brain,
  or,
• Total damage of these motor areas.

5
Consequences

• Severe difficulty with fine motor tasks (like
  writing, stitching, using computer peripherals,
  and various other such tasks.)
• Severe difficulty with any kind of communication.
• In a nutshell,
• Access to the computers is almost a dream come
  true
• The presence/absence of the peripherals are
  irrelevant for them.

6
Can Computers Help

• Certainly computers can help this population by
  being
• An easy medium of communication (which they find
  very difficult)
• An intelligent companion by understanding the
  needs and thereby reducing the communication
  efforts

7
The Impetus Something Indian!!

• Mainly the Indian scenario
• Present systems are tuned to foreign
  socio-cultural context
• All of them are imported no local support
• Costly for an average Indian user ( E Z Keys -
  1400, Gyro-HeadMouse - 1495, CameraMouseTM -
  695 costly video camera)
• Lack of Adaptation in existing systems

8
Contributions

• Implementation of a virtual adaptive mouse
  SweepSticks
• This work was a joint effort of one of my fellow
  researchers Mr. Koushik Chakraborty and myself
• Design and Implementation of the prediction
  support for a virtual keyboard (both Hindi and
  Bengali) SulekhA
• Field Testing and analysis of both SweepSticks
  and SulekhA

9
The Prelims Special Access Mechanisms

• Hardware Component Depending upon the degree of
  their motor control the disabled people can use
  either one or at most two switches (specially
  designed for them) in order to access the
  computer.

10
Special Access Mechanisms (contd)

• Software Component
• Scanning Mechanisms Guided / periodic focusing
  and defocusing of screen elements.
• Shift of focus Shift operation (needs one
  switch)
• Selection of a particular screen element
  Register operation (needs another switch)

11
Methods of Scanning
12
SulekhA A Demo
13
Learning User Preferences
14
Forming Paths
Now if the user clicks the point (30,45) once
again ???
15
Statistics

• SulekhA uses
• Bigram Prediction Strategy for Word Level
• The training corpus at present contains
  approximately 1 million words and 0.12 million
  distinct bi-grams.
• The format of the corpus is shown below,
• ltfrequency bigram1 bigram2gt
• Unigram Prediction Strategy for Character Level
• The training corpus at present contains
  approximately 1.3 million words and 0.05 million
  distinct unigrams.
• The format of the corpus is shown below,
• ltfrequency unigramgt

16
The Strategies
17
Shradha Writes with SulekhA
18
Assessments SweepSticks

• Presently tested for four subjects at IICP
  Kolkata
• Testing is done by measuring success and failure
• Success - Each user is asked to open a particular
  application using SweepSticks. If the user is
  able to do the same in one go, with the minimum
  number of shift and register operations actually
  required to open the application, then it is a
  success, else failure.
• Success and failures are measured for all the
  users for a number of sessions both in absence
  and presence of the adaptive help

19
Assessments SweepSticks
20
Assessments SulekhA

• Typing rate (number of characters typed per
  minute) was measured
• Measurements were taken when the prediction was
  not in use and also when in use

21
Usability
5 Excellent, 4 Good, 3 Average, 2
Difficult, 1 Very Difficult
22
References

• 1 Hufschmidt-Schneider M., Kuhme Thomas and
  Malinowski U.,
• Adaptive User Interfaces, Principles and
  Practice.
• 2 Ahmed Seffah and Homa Javahery, Multiple User
  Interfaces,
• Cross-Platform Applications and
  Context-Aware Interfaces.
• 3 http//www-csli.stanford.edu/cll/aui.html
• 4 http//www.words-plus.com
• 5 http//www.advancedperipheral.com
• 6 http//www.logitech.com
• 7 http//cameramouse.com
• 8 http//www.cirque.com
• 9 http//orin.com/index.htm
• 10 http//www.quadjoy.com
• 11 http//www.censusindia.net/disability/disabil
  ity_mapgallery.html
• 12 http//www.webhealthcentre.com/general/cp_in
  dia.asp
• 13 Kaul Sudha and Warrick A., Their Manner of
  Speaking, Indian
• Institute of Cerebral Palsy, Kolkata,
  India, 1997.

23
Questions