موجز عن البحث:
|
The fortunes of any language can fall and rise and are linked
familiarly to the fortunes of their speakers. Nowadays, there are many
languages becoming widely spoken. Though, it is difficult for smaller
languages to survive, a significant number of languages that have been
successfully revitalized. The purpose of this paper is to explore themes of
language death in various respects, with focus on why we should care about
languages dying. This paper also answered some questions regarding the definition
of language death and the estimated number of languages are endangered and
currently in use. Additionally, it highlighted the problems surrounding the
loss of languages, and provided some constructive solutions to reclaim and
revive dead and endangered languages. This paper also included a number of
recommendations for linguists and endangered language speakers or
communities.
|
ملخص المشاركة:
|
Although many studies have
investigated the production of reduced vowels and consonant cluster
simplification and epenthesis/vowel insertion to break up clusters, few have
taken up the opportunity to examine these in perception. We report on a study
of vowel perception by English learners from three diverse L1s. 40 Arabic-,
60 Mandarin- and 60 Spanish-speaking adolescents with eight years of
classroom English were tested on perception of schwa and the short L1 vowels
which break up disallowed consonant clusters in various environments.
Following Matthews and Brown (2004), listeners took an AX task where they had
to decide whether words matched or not in pairs of English like nonsense
words with and without vowels between consonant clusters in onsets and codas
and in medial sequences in bisyllabic words. A carrier phrase ‘I said’
ensured listeners were less likely to perceive differences when the vowel
added was at the beginning of a word and avoided acoustic processing rather
than recruitment of phonological representations as in the non-matching ‘I
said [e] stop’ ‘I said stop.’ Listeners in the three L1 groups heard five
tokens each for all epenthesis sites, with schwa and with vowels used in
Arabic (both Kuwaiti and Saudi) and in Spanish. No additional conditions were
used for Mandarin speakers since all epenthesis sites represented disallowed
consonant sequences. Corroborating Matthews and Brown, results indicated that
L1 epenthesis influenced participants’ ability to detect differences: they
better detected schwa and those vowels not used in their L1 epenthesis.
Mandarin learners were better at detecting epenthesis vowels than schwa.
Further studies are needed to address the challenge of identifying the
mechanisms driving L2 learners’ perception and in turn production of reduced
as well as short vowels.
|
ملخص المشاركة: |
Static
cues for vowel identification have been widely used, with vowel formant values
being taken at midpoint (50%) in L1 acoustics research, as it is believed that
they represent the best acoustic characteristics of monophthong vowels
(Peterson & Barney 1952). Nevertheless, many studies (e.g., Nearey &
Assmann 1986, among others) have reported that dynamic cues—in particular,
vowel-inherent spectral change (VISC) and the three-point model (e.g.,
Hillenbrand et al. 1995, among others)—can obtain better identification of
vowels by using discriminant analysis with the formant pattern sampled at
multiple locations. To date, many studies on Arab L1 vowel production have
concentrated on the static acoustic features (e.g., Almbark 2012), and the role
of dynamic cues has not been fully investigated (although see Al-Tamimi 2007).
Therefore, this research presents an acoustic description of the vowel
production of L1 Hijazi Arabic (HA) to evaluate the importance of static and
dynamic cues, particularly VISC models (direction, slope, and offset) and three
sample-points model, in the classification of HA vowels. An added purpose is to
estimate the extent to which vowel duration and F3 aid in the classification
accuracy.
Recordings were made of 12 male HA speakers producing
eight HA vowels in /hVd/ syllables that were included in a carrier sentence.
The first three formant frequencies were measured during the vowel duration
three times—once (at 50%) for the static model; twice (at 20% and 80%) for the
VISC models; and three times (at 20%, 50% and 80%) for three-point model —and
then tested using discriminant analysis. The discriminant analysis results show
that, in general, the three-point model had higher correct classification rates
for all 8 HA vowels than the static model and VISC models. However, all models
performed well in identifying HA lax and tense vowels and vowel pairs,
particularly when the vowel duration was added. The average correct
classification rate was 98% for the three-point model, 96% for the static
model, and (86% for direction, 85% for slope, and 74% for offset) for the VISC
models. The vowel duration was found to play a significant role in the
classification accuracy (+11%), while the role of F3 appeared to have a minor
influence on the classification’s accuracy (+1%). In conclusion, the
three-sample-point method was the most accurate model for classifying HA vowels
in all stages, and the vowel duration was the most important cue on the
classification accuracy of HA vowels. |