Presentation is loading. Please wait.

Presentation is loading. Please wait.

Experimenter-defined measures in a Reading Tutor that Listens

Similar presentations


Presentation on theme: "Experimenter-defined measures in a Reading Tutor that Listens"— Presentation transcript:

1 Experimenter-defined measures in a Reading Tutor that Listens
Jack Mostow Project LISTEN ( Carnegie Mellon University Funding: National Science Foundation IERI PI Meeting, Sept. 9-10, 2004 1

2 Thanks to fellow LISTENers
Tutoring: Dr. Joseph Beck, mining tutorial data Prof. Albert Corbett, cognitive tutors Prof. Rollanda O’Connor, reading Prof. Kathy Ayres, stories for children Joe Valeri, activities and interventions Becky Kennedy, linguist Listening: Dr. Mosur Ravishankar, recognizer Dr. Evandro Gouvea, acoustic training John Helman, transcriber Programmers: Andrew Cuneo, application Karen Wong, Teacher Tool Field staff: Dr. Roy Taylor Kristin Bagwell Julie Sleasman Grad students: Hao Cen, HCI Cecily Heiner, MCALL Peter Kant, Education Shanna Tellerman, ETC Plus: Advisory board Research partners DePaul UBC U. Toronto Schools 2

3 Uses of experimenter-defined measures
Analyze usage Assess students Evaluate interventions 3

4 1. Analyzing usage Participation Frequency Duration Hiatus rate
What % of enrolled students use the Reading Tutor? Frequency How often do they use it? Duration For how long at a time? Hiatus rate With how many timeouts? Crash rate With what % ending in crashes? 4

5 2002-2003 usage: lab > classroom -- but top classrooms > average lab
5

6 2. Assessing students: Latency decreases over time
Initial encounter of muttered: I’ll have to mop up all this (5630) muttered Dennis to himself but how 5 weeks later: Dennis (110) muttered oh I forgot to ask him for the money How does latency evolve in general? 6

7 Help and latency predict WI scores with correlation > 0.9.
Learning Curves for help and latency (for 2.4 million word encounters in ’00-01) Help request rate Word latency (in ms) # of previous encounters of the word Help and latency predict WI scores with correlation > 0.9. 7

8 Time as an indicator of motivation
% time picking stories predicts lower gains (R = –0.4) Mostow, J., Aist, G., Beck, J., Chalasani, R., Cuneo, A., Jia, P., & Kadaru, K. (2002, June 5-7). A La Recherche du Temps Perdu, or As Time Goes By: Where does the time go in a Reading Tutor that listens? Proceedings of the Sixth International Conference on Intelligent Tutoring Systems (ITS'2002), Biarritz, France, Hasty responses indicate guessing Beck, J. E. (2004, August 31). Using response times to model student disengagement. Proceedings of the ITS2004 Workshop on Social and Emotional Intelligence in Learning Environments, Maceio, Brazil. 8

9 3. Evaluating interventions: vocabulary [Aist PhD thesis]
Randomly explain some new words but not others. Test each new word the next day. Did kids do better on explained vs. unexplained words? Overall: NO; 38%  36%, N = 3,171 trials Rare, 1-sense words tested 1-2 days later: YES! 44% >> 26%, N = 189. 9

10 Evaluating interventions: word help [SSSR04]
Student is reading a story Student needs help on a word Tutor chooses what help to give Student continues reading Student sees word in a later sentence Time passes… ‘People sit down and …’ Student clicks ‘read.’ Decision (randomized) ‘… read a book.’ Explain outcome = word read OK, then explain segmentation artifact; read ASR as “automatic speech recognizer” ‘I love to read stories.’ Outcome: read fluently? How does outcome vary by help, word, and delay? 10

11 What helped which words best, when
What helped which words best, when? (for 270 students, 180,909 randomized trials) Best overall: Rhymes With 69.2% ± 0.4% Worst overall: Recue 55.6% ± 0.4% Compare within level to control for word difficulty. Supplying the word helped best in the short term… But rhyming hints had longer lasting benefits. Same day: Later day: Grade 1 words: Say In Context, Onset Rime Grade 2 words: Say In Context, Rhymes With Rhymes With Grade 3 words: Say In Context Rhymes With, One Grapheme 11

12 Project LISTEN’s Reading Tutor: A rich source of experimental data
database: 9 schools > 200 computers > 50,000 sessions > 1.5M tutor responses > 10M words recognized Embedded experiments Randomized trials See videos, papers, etc. at Thank you! Questions? Get hours from Joe 12

13 Project LISTEN’s Reading Tutor (video)
John Rubin (2002). The Sounds of Speech (Show 3). On Reading Rockets (Public Television series commissioned by U.S. Department of Education). Washington, DC: WETA. Available at 13

14 Map data stream to data set: trials
Context: Decision: Outcome: 14


Download ppt "Experimenter-defined measures in a Reading Tutor that Listens"

Similar presentations


Ads by Google