Communication

advertisement
Designing for authenticity: a
framework and guidelines for
conversation simulation courseware
Prof. dr. Jozef Colpaert
Frederik Cornillie
LINGUAPOLIS research & development
Language Teaching Symposium 2009,
Gent
Statement
• There’s probably no God. Now stop worrying
and enjoy your life.
Statement
• There’s probably no learning effect in
TECHNOLOGY. Now stop worrying and enjoy
your research.
Research you say ?
• NSDS: No Significant Difference Syndrome
 ecological shift: design the entire learning
environment
• SES: Sunday Evening Syndrome
– 95 % work, 5 % research
– Sunday evening = research moment?
 how to transform your daily work into research
Research you say ?
• Attitude A:
• I am involved in developing and using technology for learning and
teaching, and my perception is that I am rather good at it. I should
publish more but I have the feeling that my work is not being
valued academically nor recognized as research …
• Attitude B:
• I have just developed this tool (which already existed in the early
eighties, but who cares?). Dancing Bearware really, but I present it
as coconstruction of knowledge, negotiation of meaning, or
empirical validation of theory. It’s fake, but I add Vygotsky and it
really works. I get grants and awards, my presentations are
accepted at conferences and I even get published. You should try it
…
DLL
• Distributed Language Learning
– conceptual and methodological framework for
designing learning environments
– educational engineering model
– based on hypothetical compromise between
pedagogical and personal goals
– role of technology and choice of learning/teaching
method are logical consequences
DLL
• Distributed Language Learning
– in phase of further theoretical / empirical
validation
– contribute to collaborative research
Educational engineering
Analysis
Design
Theory
conceptualization
specification
prototyping
Development
Implementation
Evaluation
Technology
Architecture
teacher
content
content
learner
colearner
Classrm
wall
teacher
learner
colearner
Technology & authenticity
IIC model applied to learners:
Mediated
Non-mediated
Information
Interaction
Communication
…
…
…
Mediated = with human or technological intervention
Information
• Non-mediated
– use of authentic materials on the Web
– examples: WebQuests, www.deredactie.be, …
• Mediated
– use of adapted materials (simplified texts, hyperhelp …)
– examples: Internet actuel, hyperlinks,
http://podcasts.ox.ac.uk/, http://www.radiolingua.com/,
http://mitworld.mit.edu/, …
• Advantages / disadvantages ?
Communication
• Non-mediated
– With natives not aware of LL process
– Examples: chat, mail, twitter, fora, text, blog, virtual worlds
…
• Mediated
– With co-learners, teacher or natives aware of the LL
process
– Examples: Tandem learning, http://www.patati.be/
• Advantages, disadvantages ?
Interaction
• Non-mediated
– Use of systems not developed for LL purposes
– Examples: reservation sites, Second Life, (non-dedicated)
chatterbots, point of sales systems …
• Mediated
– Use of systems developed for LL purposes
– Examples: language courseware, tests,
http://www.slanguages.net/ …
• Advantages, disadvantages ?
So ?
Mediated
Non-mediated
Information
+: more adapted
-: less authenticity, labourintensive, logging
possibilities
 beginner level
+: authenticity, TBL
- : errors, quality, appropriateness,
durability …
 more advanced levels, linguistic
awareness, mature readership,
preparation and control
Interaction
+: strong didactic
functionalities
-: D&P ??
 all levels
+: authentic interaction, TBL
-: lack of tracking & feedback
 more advanced levels
Communication
+: role of coach, logging
possibilities
-: more artificial
 beginner/intermediate
levels
+: authenticity, TBL
-: errors, quality, privacy, no
coaching
 intermediate to advanced levels
Authenticity
• Different roles of technology tackle
authenticity differently
• Different moments/situations/levels require
different forms of authenticity
DISCO project
• conversation simulation
courseware for language
learning
• degrees of authenticity in ...
–
–
–
–
task type
feedback
content
assessment & progress tracking
DISCO project
• Development and Integration
of Speech technology
into Courseware for language learning
• Radboud Universiteit Nijmegen, Polderland
Language and Speech Technology, Linguapolis
(Universiteit Antwerpen)
• STEVIN research project, Nederlandse
Taalunie
DISCO requirements
• to develop and test a prototype of an ASRbased CALL application for training oral
proficiency for Dutch as a second language
(DL2) at A2 level
• to optimize learning through interaction in
realistic communication situations
• to provide intelligent feedback on various
aspects of DL2 speaking, viz. pronunciation,
morphology and syntax
Degrees of authenticity
•
•
•
•
task type
feedback
content
assessment & progress tracking
Degrees of authenticity
•
•
•
•
task type
feedback
content
assessment & progress tracking
Task type
• conversation simulation with animated agents
• “interactive participatory drama” (Hubbard
2002)
• conversation branching (item-based system)
• afterwards remediation on specific topics,
coached by pronunciation agent
Conversation environment
Conversation environment
Remediation environment
Degrees of authenticity
•
•
•
•
task type
feedback
content
assessment & progress tracking
Feedback in ASR-based CALL
• waveforms and
spectograms
• talking heads
• feedback based on
pronunciation score
• corrective feedback
Feedback in ASR-based CALL
• waveforms and
spectograms
• talking heads
• feedback based on
pronunciation score
• corrective feedback
Feedback in CAPT
Results of 2007 doctoral study, Radboud
Universiteit Nijmegen:
– The key to successful CAPT is detailed corrective
feedback ...
– ... but applied with moderation.
Dutch-CAPT project
Corrective feedback types
* Ik heb een trui gekoopt.
feedback type
classroom language learning
CALL
explicit
correction
Neen, ‘gekoopt’ is fout,
‘gekocht’ is juist.
Ik heb een trui gekoopt.
repetition of
the error
Gekoopt?
Gekoopt?
XX SEE HEIFT
recast w(/o)
stress
Oh, je hebt een trui
gek[ó/o]cht.
Oh, je hebt een trui gekocht.
clarification
request
Ik begrijp het niet. Wat bedoel
je precies?
Ik begrijp je niet. Kan je herhalen?
metalinguistic
clue
Let eens op je voltooid
deelwoord.
Het werkwoord is niet juist.
XX SEE HEIFT
elicitation
Dat is niet juist. Hoe zeggen we
dat in het Nederlands?
XX SEE HEIFT
Ambiguity of recast
• ambiguous function
– corrective & mediated (focus on form)
• “Ah, je hebt een trui gekocht.”
– communicative & authentic (focus on meaning)
• “Ah, je hebt een trui gekocht. Hoe ziet die eruit?”
• use in classroom teaching
– frequently used for treating morphosyntactic errors
– less frequently used for treating pronunciation and lexicon
• effect on learning outcome (!!)
– higher effect on pronunciation and lexicon
– lower effect on morphosyntax
Feedback in DISCO
• two types of corrective feedback, varying with
respect to learner preference
– explicit correction
• “I want lots of feedback in the conversation,
in order to correct my mistakes and achieve perfect
pronunciation.”
– recast
• “Even when I make mistakes,
I want to go on with the conversation,
in order to practice fluency. ”
• non-verbal feedback
Explicit correction
Recast
Degrees of authenticity
•
•
•
•
task type
feedback
content
assessment & progress tracking
Content
• Seemingly authentic conversations ...
– realistic situations
– speech acts
• ... however, very artificial and intricate content authoring
– technological constraints: length of sentences, interjections,
disfluencies
– phonetic constraints: frequent mistakes at A2 level
– grammatical constraints: frequent morphosyntactic mistakes at A2
level
– lexical constraints: frequency of vocabulary at A2 level
– cultural constraints: Dutch-Flemish bias
– didactic constraints: validity of assessment (see further)
=> delicate balance between requirements/constraints and
authentic feel of the conversations
Content authoring plan
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
studying the cultural context (a.o. Nieuwe Buren videos);
devising the narrative plan, with detours and alternative endings;
writing separate blocks of conversations for each alternative path in the dialogue;
feeding the conversation trees into the content database;
double-checking the logic of the narrative in the prototype software;
checking vocabulary and cultural elements for socio-linguistic consistency (Dutch-Flemish
contexts);
checking the length of the sentences from the point of view of ASR and NLP
requirements;
checking vocabulary and grammar from a linguistic-didactic point of view (typical A2
vocabulary, possible hyperonyms/hyponyms, ...);
checking vocabulary and grammar for consistency with Nieuwe Buren vocabulary and
grammar;
adding grammatical codes to each turn in the branched conversation, so that all linguistic
aspects are more or less equally distributed over the complete conversation tree;
elaborating prompts for morphology and syntax exercises (which are in fact semantically
identical to the prompts for pronunciation, but differ only in grammatical form);
adding contextual information to each pair of turns in the conversation, so that the
graphical artist and photographer know which content to create;
debugging the complete conversation in the prototype after the insertion of multimedia
content (sounds, graphics).
Content authoring in database
Content authoring in database
Degrees of authenticity
•
•
•
•
task type
feedback
content
assessment & progress tracking
Assessment & progress tracking
• formative assessment > “How am I doing?”
• system keeps scores for each topic on which learner
is tested
– phoneme
– grammatical topic (e.g. subject-verb concord, verb
position, ...)
• scores increase or decrease according to learner
performance
=> topics have to be (more or less) equally distributed
in the conversation tree
Topic distribution
Report
Summing up
XX
•
•
•
•
kort herhalen wat gezegd geweest is
IIC schema
DISCO als vb van
authenticiteit op 4 vlakken
You’re invited …
• Antwerp CALL 2010 “Beyond motivation”
– August 2010
– http://www.antwerpcall.be
• Summer School “A Toolbox for Design-Based
Research”
– Qualitative and quantitative research methods in language
learning and teaching, University of Antwerp, 16 to 20
August 2009
– http://summerschool.linguapolis.be
An authentic thank you
• Credits to
– Koen De Maesschalck, illustrations
(www.koendm.be)
– STEVIN research funds, Nederlandse Taalunie
• Contact us
– jozef.colpaert@ua.ac.be
– frederik.cornillie@ua.ac.be
Questions for discussion
• Q1
• Q2
Download