How do we correct this? • Solution: The TeNeT Online Tutorials .... Development
based on. – PHP, MySql, XML, Qt ... Speak with a Malayali accent/ Tamil Hindi ...
Online Tutorials: User Interface Design
Hema A Murthy and M Bhuvaneswari, IIT Madras Email:
[email protected] Funded partially by: Intel Technologies, Department of Science and Technology, TeNeT Royalty, IIT Madras
Purpose: Education and Training • In the absence of good schools and good teachers: – Enable students to pass the standard X examination – The target users: children • How do we design the material to keep them attentive? • How can we make the lessons effective (given that most children are I generation learners)?
Our First Experiment: Spoken English • A system with a dialogue – Beautiful pictures, animation – But NO TAKERS – Although material had rural content
• The reason – A child had to walk 3 Km (nearest kiosk) for a mere English Lesson! – The system did not interact with the child – Parents were quite unenthusiastic
Rural Children: not Ready for Spoken English • Question: Is it too elitist? • Answer: TNSF survey – Most children in rural dropout of school early • Dropout rate as high as 40% in IX and X grades – No support system to help first generation learners
• Spoken English does not take them anywhere
– How do we correct this? • Solution: The TeNeT Online Tutorials
90 80 70
Age 7 - 14
Age 7 - 10
Age 11 - 14
60 50 40 30 20 10 0 Level 1
Level 2
% of children who cannot read
(-) or (/ )
(/ )
% of chidren who can't solve numerical sums
Annual survey of education in India: Pratham
Online Education in the Vernacular: TeNeT Online Tutorials Village Internet Kiosks or Schools
Tutorials, tests, solved question papers
Online learning materials for “Rural” areas “Rural” Focusing on rural needs, demands, …
What are the Issues in developing Online Materials? • Poor infrastructure – Low bandwidth links (only 16 Kbps was available) – School does not have Internet
• The users are Children – How do we make them effective (to ensure attentiveness)? – Most children walk about 3 Km – need to maximise the learning for time spent.
Our First Lesson • Video Lectures on lessons from the curriculum – No TAKERS • Too passive
• Everybody is VARY of a COMPUTER Teacher • Lessons MUST be Interactive – Experiments on Attention Span of Children • About 22.5 minutes • No explanation should be longer than this
Our Second Lesson • Multiple Choice Questions – Radio Buttons • Children find it difficult to focus the mouse • Large buttons required
• Subjective Questions • Typing in the text box provided – finger typing • For correcting a punctuation – as long as 20 mins! • Correction became difficult – Too many different answers possible
Old Version: Radio Buttons
Old Version: Punctuation
New Version: Punctuation
Back to the Drawing Board • School teachers were brought on board – Good teachers – Teachers who correct papers
• Strategies for cracking a Question Paper – Grammar correct ==> full marks! – Map marking/Diagram marking ==> full marks! – Short/Long Questions – keywords are important ==> Guarantee 50% marks
Focus on Learning that can Guarantee Passing Exams • Which Subjects to focus – TNSF Survey: English, Maths, Science and Soc. Science (in that order)
• English – Focus as much as possible on grammar – Letter writing – focus on relevant information
• Which languages – Each state has a VERNACULAR medium of instruction ==> how do we SCALE?
Change the UI • Radio buttons converted to LIST BOXES • Encourage KEYWORD based learning – Short and long questions • Answers hover around a few keywords • Bulletise the answers – Multiple choice on keywords
– Map marking and diagrams • Use HOT keys
The Learn Session(Biology)
The Learn Session (Maps)
The Learn Session(Physics)
The Practice Session(Maths)
The Practice Session(Maps)
The Practice Session(Soc. Sc: Long Questions)
Other Issues • Assessment of students – Generation of question papers – Answer Online – Evaluate and give feedback
• How do we ensure students go to the kiosks and learn? – Incentives to study – get corporates to pay for education at kiosk
• How do we monitor lessons at the Kiosk?
Online User Login
Question Paper Generation
Evaluation of the Paper
Questions: Will this work in a REAL Exam? • Conduct a MOCK examination • 2005, 2006, 2007, 2008 – First year, children were asked to write complete answer and mark keywords. – Interestingly children marked the right keywords
Mock Examination 2006 • In January 2006 the Online Tutorial Mock Examination was conducted • It covered seven districts – Thiruvallur, Bhavani, Alagumalai, Nellikuppam, Mayiladuthurai, Orathanadu and Theni • 18 kiosks participated in the exercise • 680 students took the exam of which 158 were non Online TT students • Majority participation was from Thiruvallur, Bhavani, Theni, Mayiladuthurai and Orathanadu
Results of Mock Exam
265
257
pass
fail
Online TT pass – fail percentage (Total number of students – 522)
41
pass
117
fail
Non- Online TT pass – fail percentage (Total number of students – 158)
• Online TT students had a 50 percent (approx) pass record • Non-Online TT students had a 25 percent (approx) pass record • The huge number failures reiterated the necessity for products like Online TT that essentially aims at aiding students to pass their Board exams.
Board Exam 2006 Registrations picked up after the Mock exam 750 Online TT students took the Board Exam 93
fail
Results have come in for 580 students 487 have cleared exams bringing the pass percentage to 83.96%
487
The state pass percentage is 77.6%
pass
Performance in Mathematics and Social Science has been commendable
SSLC performance of 580 students
Students in Bhavani, Aranthangi and Theni have performed well overall Though Thiruvallur’s Board exam performance has been better than the Mock exam, marks scored by students continues to be poor
Summary of results: Public Exam • Results over the years – 2005 – 80% – 2006 – 84% – 2007 – 85% – 2008 – 89% – The results of 2008 are really significant given that the students were continuously monitored
Online TT: some snap shots
K Marudhuvendran, School Topper, Alinjivakkam
Software development – How do Scale? • The requirement • • • •
Webbased content (constraints on bandwidth) Interactive learning Bilingual content and voice overs Tools should be easy to adapt for other languages and boards • Community MUST participate in CONTENT development
• The solutions • Tools for content generation • Development based on – PHP, MySql, XML, Qt
Generate Usage Reports: Corporate Requirement
Data at Kiosks using a Word Processor
Data Entry Using the Tool
The corresponding HTML page
Where do we go from here? • Can we build speech and handwriting interfaces into the application? – Currently the system ONLY uses mouse clicks • A sort of directed learning
– Students can actually speak out/write out the answers? – Can we modify the system to enable: • Microphone, handwriting?
Screenshots
Screenshots
Screenshots
Web Interface for TTS
Issues/Challenges • Kiosk Environment – Too noisy – Too much reverberation – How do we make ASR work in SNRs as low as zero dB? – Mikes of very poor quality • How do we handle the mismatch in the environment?
Framework for Online Exams
Online Subjective Exams: Oral
Online Handwriting based Exam: Signup
Online Handwriting based Exams
Online Evaluation
Where do we go from here? • Spoken English Programmes – Use of Automatic Speech Recognition and Speech Synthesis – Development of tools for Spoken English • Directed dialogue systems? • System continuously corrects the student? • Focus not on accent but on speaking fluently – Speak with a Malayali accent/ Tamil Hindi
Spoken English Level 1
Level 1: Building vocabulary
Spoken English Level 1
When the girl is clicked the object is zoomed and
Spoken English Level 2
Level2 Building Sentences
Spoken English Level2
Level 2: The child picks words from the words given and forms a sentence. The system can read/record the spoken sentence and evaluate the spoken sentence
Where do we go from here? • Pace the learning based on the ability of the child – Machine learning techniques for education • Focus on pedagogy – Also focus on K12 program
• Speech interfaces along with machine learning can go a long way in helping with education.
Disability, ICT, Rural India • About 3% of India’s population suffers from some form of disability – A large percentage of this comes from rural areas • The tutorials have been modified to enable interaction for children with severe motor disability (ideas borrowed from HOPE)
Disability, ICT, Rural India • We need Low cost Assistive Technologies to be built in local languages – Speech recognisers, Speech synthesisers, handwriting interfaces – Communication devices » Eyepointing devices, switches
• We need JAWS like screen readers for the visually challenged – Example, SAFA (from Webel Technologies)
• Conscious attempt by scientists and engineers to make any Speech enabled tool SAPI compliant
Summary • ICTs have come to stay – Given the literacy levels in India (65%), speech interfaces can go a long way in bringing the marginalised sections into the mainstream – Technology must be inclusive – Quoting Mahatma Gandhi • “Education should be so revolutionised as to answer the wants of the poorest villager, instead of answering those of an imperial exploiter.”