Documenting and describing language
variation and change in the British
Sign Language Corpus.
Adam Schembri, Jordan Fenlon, Ramas
Rentelis & Rosemary Stamp.
“Sign Language Corpora: Linguistic Issues” workshop, 25th July 2009.
Overview
• Background to BSL Corpus Project
• Methodology
• What we are finding so far
• Phonological variation
• Lexical variation and change
Background: Aims of the BSL Corpus
Project
• To create an on-line, open-access corpus of annotated BSL digital
video data that will become a shared, peer-reviewable resource
and standard reference for BSL researchers and teachers. Full
participant consent and metadata (background data about
participants etc) will be included.
• To conduct corpus-based investigations of sociolinguistic
variation and change, language contact and lexical frequency
• Project timeline: January 2008-December 2010
Background: Content
Phase 1: 240 signers for 2 hours
• 120 filming sessions
•Warm up activity: personal
experience anecdotes
•30 minutes free conversation
•15 minute interview on Deaf
issues and language attitudes
•15 minute lexical elicitation task
• Currently underway
Phase 2: 100 native signers for
2 hours
• Elicited narratives
• Elicitation tasks to be
decided (e.g., eliciting
various key aspects of BSL
grammar), with different
tasks for different subsets of
participants
Background: Specific studies
• (1) Linguistic and sociolinguistic variation and change
in
• (a) phonological variable (1/G handshape variation)
• (b) lexical variables (100 target lexical items)
• (c) grammatical variable (agreement/indicating verbs &
subject/object drop)
• (2) Lexical frequency based on the annotation of
100,000 signs
Background:
Project team
• Research assistant (0.6,
2009-10): capture, editing,
annotation (Ramas Rentelis)
• 1 principal investigator/project • 8 Deaf community
fieldworkers (150-200 hrs
director (2008-10)
each post)
& 7 co-investigators
• 1 translator/research assistant
• Advisory Group: 9 members
(0.6, 2009-10): provide
of Deaf community
written English translation of
data (Kyra Pollitt)
• Research associate (2009-10):
data collection co-ordinator,
• 1 PhD studentship: lexical
variation and change
annotation and analysis of data
(Rosemary Stamp)
(Jordan Fenlon)
• 1 project technician (0.5,
2008-10)
Methodology: Sociolinguistic approach
•
•
Film 30 Deaf native and nearnative signers (BSL exposure by 7
years of age) in 8 regions across
the UK:
• England (London, Bristol,
Birmingham, Manchester,
Newcastle)
• Wales: Cardiff
• Scotland: Glasgow
• Northern Ireland: Belfast
Total sample of 240 individuals,
balanced for age, gender, language
background, possibly social class
and ethnicity
Methodology: Recruitment & data
collection
• Deaf community
fieldworkers (cf. ‘contact
people’) will recruit 240 (30
x 8 regions) participants that
match project criteria
• Filming over 2-3 visits
• No hearing people present
during filming
• Pairs of signers matched for
region and age
• All resident in their region
for 10 years at least
• Participants of the same or
different genders
• As few long-term
partners/spouses filmed
together as possible
Methodology: Recruitment & data
collection
•
Filming session:
• blue background screen
• two lights
• plain colored clothing (back-up T-shirts)
• chairs without arms
• 1 high definition video-camera(s) focused on each participant, 1 on the pair
Methodology: Annotation
• Annotation using ELAN template created for Auslan corpus
project
• ID glosses for 100,000 signs and for all other annotation
• Controlled vocabulary (CV) tagging for specific linguistic
variables
• Tagging each file for social factors
• Written English translation
• BSLCP team is using ELAN annotation for a study of
phonological variation
• Separate tiers for each factor group, with a summary tier with all
factors in a string for export into VARBRUL
Lexical frequency and ELAN
• BSLCP team is using ELAN annotation for a specific study of
lexical frequency
• RH and LH tier with ID-glosses: standardised for all signs and
linked to a lexical database being created
• Lexical variants distinguished by numerals: BROWN1 versus
BROWN2
• Phonological variants of a single sign (MOTHER1a and
MOTHER1b) grouped under single ID-gloss: MOTHER1
• All glosses to be exported and counted to determine most
frequent lexical items in BSL conversations: needed for
psycholinguistic studies at DCAL
Methodology: Open access archive
• 300+ hours of BSL
digital video data
• Accompanying written
English translation
• Accompanying metadata
(data about the data)
• ELAN annotation files to
be made publicly
accessible
Current status July 2009
• Data collection: 140/240
• Birmingham: 30 participants
filmed
• Glasgow: 30 participants
filmed
• Bristol: 32 participants
filmed
• Manchester: 30 participants
filmed
• London: 18 participants
filmed
• Filming in London to be
completed soon
• Fieldworker hired for
Cardiff, Belfast &
Newcastle
• 360 1.5 hour tapes of
unedited footage
So what are finding out thus far?
BSL 1 handshape variation study
•
This study replicates work on ASL (Bayley, Lucas & Rose, 2002) in the BSL signing
community
•
Variation in the handshape of BSL signs that are produced in citation form with a 1/G
handshape (e.g., PEOPLE, THINK, HEARING, QUICK, WHAT, BUT, PRO-1).
Citation form (+cf) involves straight extended index finger with no extension of thumb
or non-selected fingers. Non-citation form may involve bending of index, thumb
extension as well as extension of non-selected fingers (-cf)
•
Research questions include: Is variation in 1 handshape signs similar or different in
BSL compared to ASL? What is the role of linguistic factors in BSL, such as
grammatical category and assimilation? What social factors are relevant for this
variation in BSL?
BSL 1 handshape variation
•
•
•
•
•
Aim to collect 10 tokens from 240
participants = 2,400 tokens
Data consists of 1200 tokens coded thus far
from 120 participants in Glasgow, Bristol,
Manchester and Birmingham (i.e., this
represents only 50% of total data)
65% +citation form, 35% –citation form
Coded to test possible effects of linguistic
and social factors
Analysed using VARBRUL software
781
800
700
600
500
400
300
200
100
0
419
-cf
+cf
BSL 1 handshape variation: Factors
We coded for:
• Number of articulators in target sign: 1h vs 2h signs
• Grammatical category
•
•
content signs (e.g., THINK, PEOPLE, HEARING, QUICK)
grammatical signs (e.g., WHAT, BUT)
•
•
Handshape in the preceding and following sign:
•
•
•
•
•
•
•
pronoun signs (e.g., PRO-1, PRO-2)
1 handshape
some other handshape
no handshape due to pause in the signing
Gender (m vs f)
Age (18-50 vs 51+)
Language background (parents Deaf or hearing)
Region: Glasgow, Manchester, Bristol & Birmingham
Factor group
Factor(% of total)
% –cf (VARBRUL
weight)
Grammatical category Pronoun signs
Grammatical signs
Content signs
57.7% (.731)
26.4% (.430)
13.4%(.268)
Preceding HS
Other HS
Pause
1 HS
38.9%(.569)
36%(.434)
16.1%(.262)
Following HS
Pause
Other HS
1 HS
41.4%(.563)
39.4%(.556)
15.5% (.283)
1 vs 2 handed
1 handed
2 handed symmetrical
2 handed asymmetrical
40.8%(.558)
9.7%(.288)
4.9%(.186)
Gender
Male
Female
36.6%(.549)
33.2%(.450)
Application value -cf
c2/cell = 0.8275. All factor groups significant at p <.05. Log likelihood = 
-608.791
BSL 1 handshape study
All linguistic factors are significant:
• Like the ASL study (Bayley, Lucas & Rose, 2002),
we have evidence for assimilation of handshape due
to effects of the preceding and following sign
• Also like ASL, assimilation effects are strongest in
pronouns and grammatical signs and less common
in content signs
Only one social factor is significant:
• Unlike ASL, we also have evidence that there
women signers favour citation forms of these signs
when compared to men
• Age, region, language background not significant.
400
350
300
250
Pro
200
Gramm
150
Content
100
50
0
+cf
-cf
BSL 1/G handshape study
•
Why is this important?
• Handshape variation is not well understood in signed
languages
• The fact that pointing signs show significantly more
handshape variation than lexical signs suggests that
indexicality is interacting with linguistic form
• Is frequency important? Lexical frequency effects in the
Auslan and NZSL location variation study (Schembri et
al., 2009)
• Gender differences in spoken languages
So what are finding out thus far?
Observations on lexical variation data
• Lexical variation research questions
•
Is there evidence of dialect levelling in BSL? How does it correlate with
social factors such as region, age, gender etc?
• Preliminary observations suggest lexical variation in BSL does
appear to be diminishing: younger signers do appear to be using
more standardised lexical items than the older generation,
including number signs (e.g., Manchester numbers) and signs for
foreign countries (e.g., AMERICA)
• Dialect levelling means a loss of heritage BSL vocabulary and the
strong regional identities they represented
• BUT it is perhaps the inevitable result of the emergence of a more
national and more international Deaf identity in the UK
BSL lexical variation: Number signs in
Manchester & Birmingham
All participants asked to produce their signs for 1 to 20 in a fixed random order.
We are coding for:
1.
Each specific lexical and formational variant
2.
Whether each variants represents a traditional regional number sign or a non-traditional sign
3.
Gender (male vs female)
4.
Age (18-35, 36-50, 51-65, 66+)
5.
Language background (parents Deaf or hearing)
6.
Region: Glasgow, Manchester, Bristol, Birmingham, London, Belfast, Cardiff & Newcastle
7.
Ethnicity: White British versus Other (South Asian, Black etc)
Preliminary results, based on 1200 tokens from 60 participants in Manchester and Birmingham only,
analysed for (2), (3), (4), (5), (6) and (7)
Results: Manchester & Birmingham
Factor group
Factor
% non traditional
(VARBRUL weight)
Age
Young (18-35)
Mature (36-50)
Older (51-65)
Elderly (66+)
50.6% (.747)
29.7% (.586)
18.3%(.344)
14.7%(.254)
Region
Manchester
Birmingham
37.2%(.622)
21.9%(.378)
Language background
Hearing
Deaf
31.5%(.551)
25.3%(.392)
Application value N (non-traditional number signs)
Chi2/cell = 3.0864. All factor groups significant at p <.05. Log likelihood = 
-630.789
Results: Manchester & Birmingham
•
•
•
•
Thus, evidence of movement away from traditional numbers signs in younger
generations in both Manchester and Birmingham
Note that the difference between the two age groups 50-65 and 66+ was not
significant (Fisher’s Exact Test, p value equals 0.2569), but that the differences
between the oldest (50+), middle (36-50) and youngest groups (18-35) were
each significant (Fisher’s Exact Test, p value less than 0.0001): this suggests
that there was minimal change in the two older groups, with the two younger
age groups driving the variation
The change was stronger in the Manchester data than Birmingham, and in
Deaf adults with hearing parents than in those with Deaf parents
Gender and ethnicity were not significant
BSL Corpus Project: Acknowledgements
•
•
•
Thanks to the following researchers whose work influenced our research
design:
• Trevor Johnston (Australia)
• Ceil Lucas (USA)
• David McKee & Graeme Kennedy (New Zealand)
Thanks to the project co-investigators (Kearsy Cormier, Margaret Deuchar,
Frances Elton, Donall O’Baoill, Rachel Sutton-Spence, Graham Turner,
Bencie Woll) & Deaf Community Advisory Group members (Linda Day, Clark
Denmark, Helen Foulkes, Melinda Napier, Tessa Padden, Gary Quinn, Kate
Rowley & Lorna Allsop)
Thanks to Sally Reynolds & Annika Pabsch for data collection and editing
Contacts & websites
• Adam Schembri & Jordan Fenlon
• [email protected] & [email protected]
• DCAL Research Centre, UCL
• www.dcal.ucl.ac.uk
• Project website
• www.bslcorpusproject.org
Descargar

The Auslan archive project: who, what, where, when, how & …