Content uploaded by Tobias Haug
Author content
All content in this area was uploaded by Tobias Haug on Sep 26, 2016
Content may be subject to copyright.
!
!
!
!
!"#$$%&'(#)*'+%!,#-*,-%.)/*0'(#)*%%%%1%
2%
Summary of the International Exploratory Workshop
Methodological Challenges and Technical Innovations in Signed
Language Assessment
1. Summary of the workshop
The presentations on the first day of the workshop served as a basis for the discussions in the
three parallel sub-workshops on the second day and for the final discussion on the third day.
1.1!Summary of the presentations
Presentation 1: Key methodological issues in signed language test development for L1 and L2
learners
Robert Hoffmeister and Peter Hauser (both USA) reported on methodological issues in signed
language test development for children/L1 learners and adults/L2 learners. For example, they
talked about differences in the degree of language input that the tested deaf children had
received. Only 5% of the deaf signing population can be considered “native signers”, the remaining
95% of the children receive a variety of (signed) language input. Another issue that the presenters
stressed was the age at which deaf children receive their linguistic input, which has an impact on
their language development and consequently on their success in the school context.
As for the adult learners, the shortage of research on the acquisition of a signed language in adult
learners has an impact on test development. Little is known about the “typical” development of
adult L2 learners of a signed language. The presenters further emphasized the importance of
interdisciplinary cooperation during test development, such cooperation with psychologists,
linguists, and members of the deaf community.
Presentation 2: What are differences and similarities in signed language test design for L1 and L2
learners?
Eveline Boers (The Netherlands) and Charlotte Enns (Canada) talked about the similarities and
differences in test design for L1 and L2 learners of a signed language. One of the most striking
differences is the age of the target groups, which has an impact on the testing materials used and,
hence, on the test design. For example, images need to be age-appropriate when testing children,
whereas more complex images like a map or a calendar can be used when testing adults. Another
issue concerns the length of videos used in test tasks, i.e., a video for children should be shorter
than one for adults. Moreover, the style/register of the signer on the video needs to match the
target group. A test for children needs to be administered by an adult, whereas adults can be
tested independently. Both presenters agreed on the potential of web-based testing formats. They
also emphasized the need for future collaborations in signed language test development for both
L1 and L2 learners. For example, they mentioned that it might be possible to share elicitation
materials across signed languages.
Presentation 3: Applying a framework of computer-/mobile-assisted language testing (CALT) for
spoken languages to signed language assessment
Tobias Haug (Switzerland) and Wolfgang Mann (UK) reported on their work in applying a CALT
framework (Suvorov & Hegelheimer, 2014)1 to signed language test development. They focused on
the following aspects of the framework:
1.%Delivery format (computer- or web-based)
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
1 Suvorov, R., & Hegelheimer, V. (2014). Computer-Assisted Language Testing. In A. J. Kunnan (Ed.), The Companion to
Language Assessment (pp. 594–613). Hoboken, NJ, USA: John Wiley & Sons, Inc.
!
!
!
!
!"#$$%&'(#)*'+%!,#-*,-%.)/*0'(#)*%%%%1%
3%
2.%Media density (e.g., multimedia)
3.%Scoring mechanism (human scoring, automatic scoring)
4.%Task and response types (e.g., selected or constructed responses)
The presenters concluded that CALT offers a lot of potential for future applications in signed
language testing, potential that has not yet been used for many signed language tests. For
example, there are a number of web-based tests, but very few make use of, for example, online
video recording (media density) or apply automatic score reporting for their (multiple-choice) tests.
Also, many signed language tests often use selected-response tasks and response types, such as
multiple-choice items, and very few tests make use of productive (e.g., short answers) or
interactive (e.g., drag-and-drop) task types. The presenters emphasized that the framework they
had applied can be used as a guidepost when thinking about the possibilities that web-based
testing can offer. However, they recommended to always reflect critically on the extent to which,
for example, automatic score reporting or more interactive task types can actually improve the
authenticity, reliability and validity of a test.
Presentation 4: Automatic signed language recognition
Richard Bowden (UK) provided a comprehensive review on the development of signed language
recognition technology. Older work of signed language recognition was based on a two-stage
linguistic inspired framework where the first stage of recognition extracts high level features
similar to those used in linguistic annotation and the second stage then recognizes the temporal
order of these features. Later, Bowden discussed how this framework has progressed over the
years to include the Hamburg Notation System, a form-based annotation system for signed
languages, and modern sensors such as the Microsoft Kinect. In particular, he discussed in more
detail some of the aspects of recognition including hand shape, movement and more recent
pattern mining techniques for second stage temporal recognition. He then turned to faces and
talked about facial feature tracking, e.g., expression and non-verbal facial communication in the
context of speech including aspects of lip reading. Finally, he talked about employing mouthing's
in signed language recognition to bolster recognition performance.
!
Presentation 5: Automatic signed language generation
In the first part of their presentation, Sarah Ebling (Switzerland) and Robert Smith (Ireland) talked
about the advantages and disadvantages of different approaches to automatic signed language
generation, namely, (1) hand animation, (2) animation based on motion capturing, and (3) full
synthesis. None of the three approaches offers the perfect solution. Hand animation and
animation based on motion capturing offer high quality but are at the same time very time-
consuming and not flexible, since they consist of finite sets of signed utterances. In contrast, the
full synthesis approach offers more flexibility, as it does not consist of an inventory of pre-
determined signed utterances. The disadvantage of this approach is the lower quality of the
signing.
In the second part, Ebling and Smith reported on sign language animation acceptance and
comprehension studies with respect to the three approaches mentioned above. The hand-
animated avatars received the most positive feedback in these studies. The fully synthesized
avatars were shown to be very stiff and robot-like. Avatars have so far not been used in signed
language assessment but could offer children a more “game-like” experience. Here, the next step
would be to conduct a study that assesses the comprehension and acceptance of signing avatars
in a signed languages test.
!
!
!
!
!"#$$%&'(#)*'+%!,#-*,-%.)/*0'(#)*%%%%1%
4%
1.2!Summary from the sub-workshops
On the second day, three parallel sub-workshops were offered. Each sub-workshop was provided
twice, i.e., each workshop participant could visit two different sub-workshops. The sub-workshops
dealt with the following topics:
1.%Methodological issues in signed language test development
2.%Computer-/mobile-assisted language testing (CALT) for signed language assessment and
3.%Consequences of signed language testing for the learners
Summary of sub-workshop 1: Methodological issues in signed language test development
The following topics were discussed that will also inform the guidelines and have been used for
the research agenda:
1.%Qualifications of and training for test administrators and raters (for example, high
proficiency in a signed language, test administration).
2.%Qualifications of people involved in signed language test development: deaf people,
practitioners, researchers, media design people, programmers etc.
3.%Apply different tasks/items to assess the signed language development of L1 and L2
learners, such as vocabulary tests, judgment of sentence correctness, narrative tasks.
4.%Experiment with task and response types that have not yet been used in signed language
tests, such as cloze procedure, gap filling, or combining sentences.
5.%Testing should be more tightly integrated with the curriculum (relating to the testing of
deaf children).
Summary of sub-workshop 2: Computer-/mobile-assisted language testing (CALT) for signed
language assessment
The following issues related to CALT, automatic signed language recognition, and generation were
discussed in this sub-workshop and will also inform the guidelines and has been used for the
research agenda:
1.%Automatic signed language recognition and the use of avatars offer the possibility of
anonymization for the test takers, both children and adults. Deaf communities are known
to be rather small, therefore this possibility could increase, for example, intra-rater
reliability.
2.%Use of automatic signed language recognition systems for automatic scoring in tests of
productive signing, i.e., the recognition system rates the production of the test taker. Still,
the results would need to be checked by human raters due to the variability that is
acceptable, for example, in a sentence repetition test.
3.%Internet bandwidth poses a problem for the use of videos, maybe use vimeo or other
services for minimizing the problem of different video formats.
4.%General technical issues, such as server connectivity, video formats and file size, quality of
webcams; provide good-practice examples in the guidelines.
5.%Signed language tests should be developed by a team that contributes expertise from
different areas, such as the technical side, but also expertise on signed languages and test
construction.
6.%Investigate to what extent interaction can be assessed using, for example eye-tracking
technology.
7.%Use of a signed language assessment portal that can be used across signed languages, by
different user groups such as practitioners, researchers, test takers, test administrators.
This portal should also include the possibility to provide feedback to the learner.
8.%Test developers need to get a clear understanding of what is technically possible; this
should inform the guidelines.
9.%Gamification, especially for the testing of children.
!
!
!
!
!"#$$%&'(#)*'+%!,#-*,-%.)/*0'(#)*%%%%1%
5%
Summary of sub-workshop 3: Consequences of signed language testing for learners
The consequences for children as L1 learners are different (e.g., start with an intervention) than
for adult L2 learners of a signed language (e.g., passing/failing a final exam). The following issues
were considered as important, separately for L1 and L2 learners:
Consequences for L1 learners:
1.%Consequences of an intervention for a deaf child, or placement in a different class/unit
within a school
2.%Very few signed language tests are available worldwide, most of them do not account for
the different sub-groups of deaf signed language learners.
3.%The need for signed language tests to form an established part of schools for the deaf that
work bilingually
4.%Availability of L1 tests and information about parents would affect educational policies.
5.%Lack of access to a language (spoken and/or signed) has an impact on deaf children’s
lives.
Consequences for L2 learners:
1.%Impact of consequences for adult L2 learners, i.e., what are the stakes, for example, for
teachers whose job may depend on the test results or a promotion.
2.%Higher level of language proficiency according to the Common European Framework of
Reference is often lacking, both for teaching and learning, but also for assessment.
3.%Self-assessment should be used more often to support signed language learning.
4.%The importance of providing feedback on how signed language proficiency can be improved
to the learners; importance of gaining feedback on signed language competencies in, for
example, signed language interpreting students from the local deaf community
5.%Development of tests that tap into the linguistic skills and world knowledge of the
language learners, i.e., the question of whether we are testing language or other
skills/abilities.
6.%Develop course materials and assessment procedures for hearing parents of deaf children
who start learning a signed language.
The results of this sub-workshop will contribute to an ongoing debate on consequences of signed
language assessment, which has not yet been addressed in the literature.
1.3!Outputs: Results of final discussion
It was decided at the workshop to set up an international steering group that will contribute to
the guidelines. Eight workshop participants are interested in joining this steering group,
consisting of deaf and hearing experts, with background in testing children or adults. The
interested persons will meet via Skype in January 2016.
In what follows, the aims of the Exploratory Workshop will be revisited and it will be explained
how they have been achieved. Minutes were taken during the final discussion, which will feed into
the guidelines and will be distributed in January to all workshop participants. The guideline will
be made available as laid out in the proposal and be submitted to a journal for publication.
Aim!1!
6)% develop! and! disseminate! methodological! guidelines%7)8% (9-% 0-:-+);<-*(% )7% $#=*-0% +'*=/'=-% (-$($% >?%
,)<>#*#*=%,/((#*=@-0=-%8-$-'8,9%'*0%-A;-8(#$-%)7%'*%#*(-80#$,#;+#*'8?%(-'<%)7%>)(9%-A;-8#-*,-0%'*0%$('8(#*=%0-'7%
'*0%9-'8#*=%8-$-'8,9-8$%#*:)+:-0%#*%$#=*-0%+'*=/'=-%(-$(%0-:-+);<-*(B%
%
During the final discussion it was decided that, among others, the following topics needed to be
included in the guidelines:
!
!
!
!
!"#$$%&'(#)*'+%!,#-*,-%.)/*0'(#)*%%%%1%
C%
1)%Check existing guidelines as to whether they can be applied to signed language test
development, for example, the “Assessment Guidelines” of the European Forum of Sign
Language Interpreters (efsli, 2013)2 and different guidelines from the International Testing
Commission3. Also check the ethical guidelines from the Sign Language Linguistics Society4.
2)%The importance of having deaf people involved in teams developing tests (from development to
dissemination), e.g., deaf native signers with academic training or deaf people that can be
trained.
3)%Having people with different areas of expertise involved in the development of a test, such as
linguists, psychologists, high-level interpreters, media design experts, computer programmers.
4)%The guidelines should be available in English and International Sign.
5)%Include results from corpus projects to inform linguistic variation and change of signed
languages.
6)%The test format should match the research questions and the aims of a test’s target group
(e.g., size of population, language, heterogeneity, usage).
7)%Technical issues: provide good-practice examples for video formats etc.
8)%Ethical issues, e.g.,
a)%Approval of test taker (or test taker’s legal representative)
b)%Data have to be encrypted when they are put on a web-based signed language portal (e.g.,
“SSL certificate”).
9)%Rely on most recent research (e.g., linguistic theory) when developing a test.
10)%Maximize test usability.
11)%Community approval (face validity).
%
Aim!2!
6)% #0-*(#7?% D-?% #$$/-$% #*% $#=*-0% +'*=/'=-% (-$(% 0-:-+);<-*(% (9'(% ,'*% 9-+;% ()% 7)8<% (9-% >'$#$% 7)8% '*%
interdisciplinary! and! international! research! agenda.%69-$-% D-?% #$$/-$% "#++% ,+/$(-8%'8)/*0% (9-% 7)++)"#*=% (")%
0)<'#*$E% F'G% <-(9)0)+)=#,'+% #$$/-$% #*% $#=*-0% +'*=/'=-% (-$(% 0-:-+);<-*(% '*0% F>G% $-+-,(-0% '$;-,($% )7% HIJ6%
F0-+#:-8?%7)8<'($K%8-$;)*$-%(?;-K%('$D%(?;-K%<-0#'%0-*$#(?K%$,)8#*=%<-,9'*#$<G%';;+#-0%()%$#=*-0%+'*=/'=-$B!
!
Workshop participants discussed the following topics to be included in the international research
agenda:
1.%Usability studies on different types of response and task formats
2.%Use of a web-based portal for signed language tests across countries that can serve
different purposes (e.g., application, research).
3.%Acceptance studies of avatar-generated test items
4.%Use of different types of tests, such as more objective (e.g., multiple-choice) or subjective
tests (e.g., interview form) and their impact on test performance
5.%Use of self-assessment in adult learners and how self-assessment enhances the learning of
a signed language.
6.%Approach testing of signed interaction
%
Aim!3!
6)% discuss! consequences! of! signed! language! testing! for! the! learners%F-B=BK% ()% $('8(% $#=*-0% +'*=/'=-%
#*(-8:-*(#)*% "#(9% 0-'7% ,9#+08-*L% ,)*$-M/-*,-$% )7% (-$(% 8-$/+($% 7)8% '0/+($% J3% +-'8*-8$% #*% '*% #*(-8;8-(-8% (8'#*#*=%
;8)=8'<GB%
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
2 European Forum of Sign Language Interpreters. (2013). Assessment guidelines for sign language interpreting training
programmes.
3 https://www.intestcom.org/page/5
4 http://slls.eu/slls-ethics-statement-for-discussion/
!
!
!
!
!"#$$%&'(#)*'+%!,#-*,-%.)/*0'(#)*%%%%1%
N%
%
The consequences of signed language testing for the learners have been discussed and
documented in the minutes of the workshop. The minutes will be made available to all workshop
participants in January. Some workshop participants were interested in following up on this topic
in the form of a publication.
%
Aim!4!
6)%$(8-*=(9-*%international!research%,)++'>)8'(#)*%'*0%*-(")8D$B%
%
There were many opportunities for the workshop participants to interact with each other and
strengthen collaboration and networks.
1.4!Milestones after the workshop
All but one milestones were achieved (see box below). The milestone that is still pending
achievement concerns the creation of guidelines. The guidelines are expected to be ready at the
end of February 2016. They will be sent to the SNSF.
Milestones!
•%!-;(-<>-8%2OK%3P2CE%!/><#$$#)*%)7%8-$/+($%)7%$/>@")8D$9);$%()%")8D$9);%)8='*#Q-8%F>?%")8D$9);%,9'#8$GB%
•%!-;(-<>-8%2OK%3P2CE%!-((#*=%/;%'%,)8-%=8)/;%()%7#*'+#Q-%(9-%=/#0-+#*-$%'*0%8-$-'8,9%'=-*0'B%
•%R,()>-8%2NK% 3P2CE% .#*'+#Q#*=% =/#0-+#*-$% '*0% 8-$-'8,9% '=-*0'% '*0% $-*0#*=% )/(% 7--0>',D%()%")8D$9);%;'8(#,#;'*($%
F'++)"#*=%(")%"--D$%(#<-GB%
•%R,()>-8%4PK%3P2CE%S*,)8;)8'(#*=%7--0>',D%)7% ")8D$9);%;'8(#,#;'*($%#*%7#*'+% :-8$#)*%)7%(9-% =/#0-+#*-$%'*0% 8-$-'8,9%
'=-*0'K%;8-;'8#*=%7)8%0#$$-<#*'(#)*B%
•%&):-<>-8%3PK%3P2CE%T#$$-<#*'(#)*%)7%7#*'+%0),/<-*($B%
2. Future collaboration
During the workshop, participants had the possibility to network and discuss future
collaborations. The following collaborations emerged:
1.%A usability study on avatar-generated test tasks.
2.%Linking dynamic assessment to the learning of adult learners of a signed language.
3.%Applying an interview-like test to children.
3. Interpretation
The interpretation English/International Sign contributed greatly to the accessibility of this event
for the deaf workshop participants and secured the interaction between them and their hearing
peers, which benefited both sides.
!
!
!
!
!"#$$%&'(#)*'+%!,#-*,-%.)/*0'(#)*%%%%1%
U%
4. Appendix: Final program
Day 1: September 4, 2015
Time
Title
Presenter(s)
9.00-9.30
Welcome note
Tobias Haug
9.30-10.30
Key methodological issues in signed
language test development for L1 and
L2 learners (aim 1 & 2)
Robert Hoffmeister (L1) & Peter
Hauser (L2)
10.30-11.00:
Coffee break
11.00-12.00
What are differences and similarities in
signed language test design for L1 and
L2 learners?
Charlotte Enns (L1) & Eveline
Boers (L2)
12.00-13.30:
Lunch
13.30-14.30
Applying a framework of computer-
/mobile-assisted language testing
(CALT) for spoken languages to signed
language assessment (aims 2 & 4)
Tobias Haug & Wolfgang Mann
14.30-15.30
Automatic signed language recognition
(aim 2)
Richard Bowden
15.30-16.00:
Coffee break
16.00-17.00
Automatic signed language generation
(aim 2)
Sarah Ebling & Robert Smith
17.00-17.15h
Summary of the day
Jörg Keller
Dinner, self
organized
Day 2: September 5, 2015
Time
Title
Presenter(s) / Workshop
facilitator(s)
9.00-9.30
Information from the organizer and
information about the sub-workshops
Tobias Haug
Parallel sub-
workshops, first
round
9.30-11.00
Sub-Workshop 1:
Methodological issues in signed
language test development
Krister Schönström & Peter
Hauser (in International Sign
only)
Sub-Workshop 2:
Computer-/mobile-assisted language
testing (CALT) for signed language
assessment
Wolfgang Mann & Tobias Haug
(with interpreters)
Sub-Workshop 3:
Consequences of signed language
testing for the learners
Beppie van den Bogaerde &
Lorraine Leeson (with
interpreters)
11.00-11.30:
Coffee break
11.30-12.30
Continuation of sub-workshops 1-3
12.30-14.00:
Lunch
Parallel sub-
workshops,
second round
!
!
!
!
!"#$$%&'(#)*'+%!,#-*,-%.)/*0'(#)*%%%%1%
O%
14.00-15.30
Sub-Workshop 1:
Methodological issues in signed
language test development
Annieck van den Broek &
Evelien Boers (with interpreters)
Sub-Workshop 2:
Computer-/mobile-assisted language
testing (CALT) for signed language
assessment
Jon Henner & Patrick
Boudreault (in International
Sign)
Sub-Workshop 3:
Consequences of signed language
testing for the learners
Geoff Poor & Rosalind Herman
(with interpreters)
15.30-16.00:
Coffee break
16.00-17.00
Continuation with sub-workshops 1-3
17.00-17.30
Summary of the day/wrap-up
Tobias Haug & Wolfgang Mann
17.30-ca. 20.00
Apéro riche
(“dinner”)
Day 3: September 6, 2015
Time
Title
Presenter(s)
9.00-9.15
Information from the organizer
Tobias Haug
9.15-9.45
Summary of sub-workshops 1
Sub-workshop participant(s)
9.45-10.15
Summary of sub-workshop 2
Sub-workshop participant(s)
10.15-10.45:
Coffee break
10.45-11.15
Summary of sub-workshop 3
Sub-workshop participant(s)
11.00-13.00
Final discussion and decisions