Schedule: 2009-09-11 (16:30 - 17:30)
Poster Presentations and Commercial Exhibition (Exhibition Hall)

Title: Simulated conversation training using the Glexa software suite
Authors: Goh Kawai
Abstract: We developed Glexa, a suite of software modules for specific language
learning tasks. Our Eurocall poster describes two modules for simulated
conversation training: GlexaPhone and GlexaScene. Glexaphone is an automated web-based system for practicing telephone calls. GlexaPhone plays pre-recorded prompts in semi-random order to simulate actual conversations. To provide students and instructors with
high-fidelity audio, we initially developed GlexaPhone to run locally on
each student's computer [1]. We then improved it to run over the web
[2]. Instead of streamed audio which may sputter, audio files are
losslessly downloaded and uploaded between the client and the server.

Over 2600 students per year at Hokkaido University use GlexaPhone as
part of an online English language course required of all freshmen [3].
Students quickly become aware of the appropriate pairing between system
query and student response. Students need to immediately respond to
spoken questions. The system does not record late responses. Students
verify they answered correctly by listening to their recordings. In
2008, although 88 percent of the students had never made a phone call in
English, 97 percent succeeded after self-paced learning lasting about an
hour (we verified all phone calls by listening to all recordings). The
fact that students could judge their own performance contributed to the
high success rate, which, in turn, means that calls may not need to be
verified and graded. (We do not grade pronunciation quality, however.)
89 percent of the students recommended that GlexaPhone be used in next
year's online course.

In one of our conversations, only 17 prompts were recorded, but because
the prompts are combined in random order, students are connected to over
2,000 versions of the same call. The large variation prevents memorizing
the assignment. We believe that students build automatic reflexes to
spoken language. GlexaScene is a dialog trainer suited for pair-work in smaller, instructor-led courses. A comic-strip style graphical interface
featuring two participants (e.g., a nurse and a patient) and their lines
(e.g., "I'd like to see a doctor") starts by asking a pair of students
to say each part one by one. Each spoken part can be individually and
repeatedly recorded. Students choose their best "takes" and concatenate
them to play back what might be their conversation under ideal
conditions. In 2007, 50 nursing students used GlexaScene to practice and learn a
dialog between a patient and medical practioners, starting at the
patient's arrival at the clinic, followed by triage, history, x-ray,
treatment plan, paying the bill, and filling a prescription. All of our
students (who are not known for their language proficiency) successfully
performed the entire conversation in their midterm examination. Student
satisfaction is evidenced by this course's subsequently receiving a
first-place award in teaching excellence.

REFERENCES

[1] Goh Kawai, Yasushi Kawai and Akio Ohnishi "The design and field test
of Voice Chat, an asynchronous tool for exchanging voice messages"
Proceedings of the Acoustical Society of Japan 2005 Fall Conference,
Sendai, Japan (2005-09-27)

[2] Akio Ohnishi and Goh Kawai "A computerized telephone conversation
simulator for listening and speaking practice" CALICO 2007, San Marcos,
Texas, USA (2007-05-24)

[3] Goh Kawai "Large-scale online English language learning at Hokkaido
University" Sapporo Gakuin University CALL Workshop 2007, Sapporo, Japan
(2007-10-20)

[4] Goh Kawai and Akio Ohnishi "GlexaMotion, an audiovisual
comprehension learning experience" WorldCALL 2008, Fukuoka, Japan
(2008-08-06)
Keywords: Mobile devices, language testing

Main topic: Mobile Learning

Biodata: Goh Kawai has a BA in Engineering (University of Tokyo), a MA (International Christian University) and a PhD in linguistics (University of Tokyo). He has also been awarded as both teacher and researcher and is currently a visiting scholar at the University of Antwerp (Belgium). His current interests are voice recognition and software architecture. Among his most exciting projects is the GLEXAPHONE project for telephone and online voice recognition. He has an extensive list of publications including papers in the Journal of the Acoustical Society of Japan, Japan Educational Technology Journal or Journal of the Japan Artificial Intelligence Society. Besides he loves piloting, the Tea ceremony, amateur radio and scuba diving.
Type of presentation Poster

Paper category (Other)

Target educational sector Higher education

Language of delivery English

EU-funded project No

Schedule:	2009-09-11 (16:30 - 17:30) Poster Presentations and Commercial Exhibition (Exhibition Hall)
Title:	Simulated conversation training using the Glexa software suite
Authors:	Goh Kawai
Abstract:	We developed Glexa, a suite of software modules for specific language learning tasks. Our Eurocall poster describes two modules for simulated conversation training: GlexaPhone and GlexaScene. Glexaphone is an automated web-based system for practicing telephone calls. GlexaPhone plays pre-recorded prompts in semi-random order to simulate actual conversations. To provide students and instructors with high-fidelity audio, we initially developed GlexaPhone to run locally on each student's computer [1]. We then improved it to run over the web [2]. Instead of streamed audio which may sputter, audio files are losslessly downloaded and uploaded between the client and the server. Over 2600 students per year at Hokkaido University use GlexaPhone as part of an online English language course required of all freshmen [3]. Students quickly become aware of the appropriate pairing between system query and student response. Students need to immediately respond to spoken questions. The system does not record late responses. Students verify they answered correctly by listening to their recordings. In 2008, although 88 percent of the students had never made a phone call in English, 97 percent succeeded after self-paced learning lasting about an hour (we verified all phone calls by listening to all recordings). The fact that students could judge their own performance contributed to the high success rate, which, in turn, means that calls may not need to be verified and graded. (We do not grade pronunciation quality, however.) 89 percent of the students recommended that GlexaPhone be used in next year's online course. In one of our conversations, only 17 prompts were recorded, but because the prompts are combined in random order, students are connected to over 2,000 versions of the same call. The large variation prevents memorizing the assignment. We believe that students build automatic reflexes to spoken language. GlexaScene is a dialog trainer suited for pair-work in smaller, instructor-led courses. A comic-strip style graphical interface featuring two participants (e.g., a nurse and a patient) and their lines (e.g., "I'd like to see a doctor") starts by asking a pair of students to say each part one by one. Each spoken part can be individually and repeatedly recorded. Students choose their best "takes" and concatenate them to play back what might be their conversation under ideal conditions. In 2007, 50 nursing students used GlexaScene to practice and learn a dialog between a patient and medical practioners, starting at the patient's arrival at the clinic, followed by triage, history, x-ray, treatment plan, paying the bill, and filling a prescription. All of our students (who are not known for their language proficiency) successfully performed the entire conversation in their midterm examination. Student satisfaction is evidenced by this course's subsequently receiving a first-place award in teaching excellence. REFERENCES [1] Goh Kawai, Yasushi Kawai and Akio Ohnishi "The design and field test of Voice Chat, an asynchronous tool for exchanging voice messages" Proceedings of the Acoustical Society of Japan 2005 Fall Conference, Sendai, Japan (2005-09-27) [2] Akio Ohnishi and Goh Kawai "A computerized telephone conversation simulator for listening and speaking practice" CALICO 2007, San Marcos, Texas, USA (2007-05-24) [3] Goh Kawai "Large-scale online English language learning at Hokkaido University" Sapporo Gakuin University CALL Workshop 2007, Sapporo, Japan (2007-10-20) [4] Goh Kawai and Akio Ohnishi "GlexaMotion, an audiovisual comprehension learning experience" WorldCALL 2008, Fukuoka, Japan (2008-08-06)
Keywords:	Mobile devices, language testing
Main topic:	Mobile Learning
Biodata:	Goh Kawai has a BA in Engineering (University of Tokyo), a MA (International Christian University) and a PhD in linguistics (University of Tokyo). He has also been awarded as both teacher and researcher and is currently a visiting scholar at the University of Antwerp (Belgium). His current interests are voice recognition and software architecture. Among his most exciting projects is the GLEXAPHONE project for telephone and online voice recognition. He has an extensive list of publications including papers in the Journal of the Acoustical Society of Japan, Japan Educational Technology Journal or Journal of the Japan Artificial Intelligence Society. Besides he loves piloting, the Tea ceremony, amateur radio and scuba diving.
Type of presentation	Poster
Paper category	(Other)
Target educational sector	Higher education
Language of delivery	English
EU-funded project	No