The Hyderabad Experiment

1
The Hyderabad Experiment
September - December 2002
Sugata Mitra, CRCS, NIIT, 2002
2
Hypothesis
Given the appropriate digital resources, groups
of children can improve their English
pronounciation, by themselves....
3
Location
Peace High School Edi Bazaar Hyderabad
4
(No Transcript)
5
Experimental Setup
- Multimedia computer, freely accessible by
children - English training software -
Speech-to-text engine - Popular English films
.... No instructions
6
Procedure
The system was trained to recognise
four "standard" voices
The Kendall Coefficient of Concordance was used
for ranking
7
Observations
Children found using all features of the
system....
8
(No Transcript)
9
(No Transcript)
10
Results
September - November 2002
11
Day 0
12
(No Transcript)
13
After 10 hours of practice spread over five
months....
14
(No Transcript)
15
Day 0
16
(No Transcript)
17
After 10 hours of practice spread over five
months....
18
(No Transcript)
19
Table 1 Kendalls Coefficient of Concordance (W)
20
Measurements of improvement
21
Conclusions
A speech-to-text program can be as good a judge
of pronunciation as a group of human
judges. Groups of children can use feedback
from a speech-to-text program to improve
their pronunciation.