Professional Documents
Culture Documents
http://acousticalsociety.org/
Rosenhouse et al.
INTRODUCTION
Arabic and its dialects
Arabic is a Semitic language with a long history. Currently, it is usually divided into two registers or varieties: Modern
Standard Arabic (MSA), which is the formal, written register, and Colloquial dialects (CA), which are informal and mostly
spoken (though sometimes also written). CAI reflects here colloquial Arabic varieties spoken in Israel. Phonetic systems
of CAI (and most other CA dialects) somewhat differ from MSA. For example, Arabic MSA and CAI have short and long
vowels but in MSA there are three pairs of vowels: / i:, u:, a:, i, u, a/, whereas CAI has five pairs: /i:, e:, u:, o:, a:, i, e,
u, o, a/.1
Since Arabic dialects vary, they are usually classified by the following categories:
Geography: northern, central and southern dialects
Demography: Sedentary (urban vs. rural) and nomadic (Bedouin) dialects
Social classification: by religion, gender, age, etc.
Education (schooling, literacy): affects CAI (Kaye & Rosenhouse, 1997)
This paper aims to study the vowels in two CAI dialects and compares them with other Arabic dialects and with some other
languages.
on the page.
( ) ____
Method: The participants were recorded separately in a quiet room after reading the test words, to avoid pronunciation
errors. Then formants, durations and F0 (the latter is not reported here) were analyzed using Praat (Boersma, 2001) and
MATLAB (2010). Formant calculation was based on LPC algorithm (Rabiner & Schafer, 1978) with pre-emphasis at 50
Hz. To eliminate Formant calculation errors (e.g., Escudero et al., 2009), the sound files were down-sampled to 8 kHz prior
to analysis, and then LPC was applied with a user-specified LPC model order. For consistency, formants for each vowel
We do not analyze the front and a back allophones of /a, a:/, as they are not phonemic.
Rosenhouse et al.
were calculated at the highest model order achievable without "formant splitting", i.e., without obtaining two
resonances in the LPC model representing a single formant (Amir et al., submitted).
FINDINGS
CAI results
The CAI results for the MD and GD speakers are shown in Fig. 1 in scatter plots and error ellipses of F1-F2 vowel space
and in Fig. 2 for the schematic vowel trapezes for each group of speakers (men, women, Galilee, Muthallath). Fig. 3
presents the duration features of the two dialects pooled together.
List B
List C
Arabic word
gloss
Arabic word
gloss
Arabic word
gloss
bi:r
water well
fi:l
elephant
MD ri:
wind
GD: ki:s
bag
MD: fe:n
fan (hair)
GD: we:n
where
na:r
mo:z
tu:t
se:f
sword
ze:t
oil
fire
da:r
house, home
ba:b
door
bananas
MD: bo:t
shoe
mo:t
death
GD: lo:z
almonds
fu:l
broad beans
MD: mu:s
knife
GD: u:t
fish
seeds
strawberries
MD: ridel
Leg
GD: ider
leg
bat(t)
ducks
miter
meter
Bizer
MD: med(d)
stretch
MD: deb
jeep
GD: ed(d)
count!
GD: ed(d)
bes(s)
cat
raf(f)
shelf
Sad(d)
blocked
dob(b)
bear
dob(b)
bear
rod(d)
answer!
sufun
ships
mudon
towns
MD: furun
oven
GD: furon
oven
Rosenhouse et al.
FIGURE 1. Scatter plots and error ellipses of the F1-F2 vowel space for both dialects and both genders
FIGURE 2. The vowel spaces for each dialect and gender (Left: MD. Right: GD, men: top, women; bottom)
Rosenhouse et al.
a: Mean duration and STD of short (/i e a o u/) and long (/i: e: a: o: u:/) vowels. Gray bars = short, white bars = long
vowels; b: Duration ratios and STD of the short/long vowel pairs /i-i:, e-e:, a-a:, o-o:, u-u:/).
These findings reveal Inter-dialect and inter-gender acoustical differences between the two dialects. These findings can
be summarized as follows:
F1 results in MD & GD
F1 of /i/ is equal to /i:/ in MD (but not in GD)
F1 of/e/ is equal to /e:/ in GD (but not in MD)
F1 of /a/ is not equal to /a:/ for both men and women in MD (but not in GD).
F1 of /o/ is equal to /o:/ in GD and MD.
F1 of /u/ is not equal to /u:/ in both dialects
The overall F1 patterns are similar for men and women of the same dialect (see Figures 1, 2 above).
F2 results in MD & GD
The contrast pattern is nearly identical for men and women. For short vowels, three distinct tongue-position categories
emerged as expected: front (/i, e/), mid (/a/), and back (/o, u/). For long vowels, 4 distinct tongue-position categories
emerged: Front (/i:/), mid-front (/e:/), mid-back (/a:/), and back (/o:, u:/).
Vowel duration in MD & GD
Long vowels have longer durations than short vowels. Long to short vowel duration ratio ranges between 1.7 and 2.2.
Rosenhouse et al.
Table 2. The variety in vowels of eight Arab countries dialects from Abou Haidar (1994) (Cf. Newman and Verhoeven, 2002).
Values in bold are the highest in range, those underscored are the lowest.
_________________________________________________________________
vowel
i:
formant F1
Qatar
i
F2
u:
F1 F2
F1 F2
u
F1 F2
a:
F1 F2
F1
F2
310 1990 500 1400 310 830 490 1005 621 1280 620 1540
Lebanon 280 2010 490 1530 330 795 475 1060 610 1430 640 1390
Saudia
305 2530 540 1830 375 930 540 1190 730 1540 695 1590
Tunisia 315 2275 510 1690 360 830 540 1135 610 1780 650 1590
Syria
330 2465 415 2135 320 620 430 1200 710 1560 700 1680
Sudan
325 2220 420 2000 380 900 455 1040 730 1500 660 1600
UAE
335 2065 460 1720 350 990 475 1075 730 1380 640 1660
Jordan
320 2295 565 1720 260 795 580 1240 770 1521 780 1620
___________________________________________________________________
_________________________________________________________________
vowel
formant
i:
F1
i
F2
F1
u:
F2
F1 F2
a:
F1 F2
F1 F2
a
F1 F2
Lebanon1 280 2010 490 1530 330 795 475 1060 610 1430 640 1390
Syria1
330 2465 415 2135 320 620 430 1200 710 1560 700 1680
Jordan1
320 2295 565 1720 260 795 580 1240 770 1521 780 1620
MD men
375 1931 385 1713 391 1023 418 1096 591 1296 551 1360
Md women 456 2345 458 2068 456 1119 498 1335 770 1541 699 1621
Gd men
361 2013 391 1765 382 965 447 1183 597 1270 568 1253
GD women 411 2416 443 2180 444 1086 496 1405 728 1593 697 1523
__________________________________________________________________
Rosenhouse et al.
Table 4. Summary of Jordanian, Lebanese & Syrian vowel formants compared to MD and GD
Vowel
Description
F1, F2
i:
F1
F2
F1
F2
Male CAI values < Jordanian CA; female CAI values almost as Syrian CA values
U: F1
F2
U F1
F2
A: F1
F2
A F1
F2
Male CAI about as high as Syrian CA; female CAI - between Lebanese and Jordanian CA
Male CAI values are between Lebanese and Syrian CA; female CAI values are > all three Levant dialects
Male CAI < the three CA dialects; female CAI values between Syrian and Jordanian CA
Male CAI values are < all Levant values; female CAI values > Syrian CA (the highest of the 3 dialects)
Male CAI < than the three CA dialects; female CAI values almost as Syrian CA
Male CAI values are < all Levant values; female values are between Lebanese and Jordanian CA
We further compared our MD and GD findings with those given in Al-Tamimi (2002), which compared Moroccan (MA)
and Jordanian (JA) vowels. The present comparison reveals that like MD and GD, JA and MA production of male vowels
shows a vowel space which is more centralized than the female vowel space (p<0.0001). In addition, like MD and GD, JA
and MA short vowels are more centralized than long ones. (This is also true for the Palestinian dialect studied by Saadah,
2011). Al-Tamimi (2002) also found that JA (but not MA) speakers produce/ i/ and /e:/ with larger distinction than they
perceive them. This observation for a dialect is similar to our CAI findings. This also suggests that at least the status of
these vowels differs from the status of vowels in more distant CA dialects. In addition, JA production revealed, like CAI,
four tongue positions.2
VOWEL SYSTEMS:
EXMAPLES OF QUANTITY LANGUAGES AND NON-QUANTITY LANGUAGES
As well known, vowel systems vary in different languages. Some languages have parallel short/long vowel pairs; others
do not have symmetrical structures; and a member of a duration category (short/ long) may differ from the other member.
Unlike Al-Tamimi, we did not study vowel perception in CAI, but MD male and female /i, e/ production is very close and comparable
to Al-Tamimi's JA vowel space; cf. illustration 12 in Al-Tamimi (2002), and Figure 1 and 2 above.
Rosenhouse et al.
The figures below show vowel spaces of five quantity languages (English, German, French, Swedish and Hungarian) and
a non-quantity language (Modern Hebrew) which has five vowels, like CAI.3
Hebrew (Amir & al., 2012) Left: men, Right: women; red: stressed, blue: unstressed; black: Most et al. 2000.
FIGURE 4. Vowel spaces of five quantity languages and a non-quantity language
The Spanish vowel trapeze is very similar to the Hebrew one and therefore is not presented here.
Rosenhouse et al.
Dispersion Theory (DT), among other phonological theories, studies vowel arrays (Ferrari-Disner, 1983, Becker-Kristal,
2010 Vaux, 2010). According to this theory, vowel systems can be small or large, symmetric or asymmetric, peripheric or
non-peripheric etc. Small systems have usually less than eight vowels. The larger the system, the more categories it usually
has (short/long, symmetric/ asymmetric, peripheric/non-peripheric, etc.). Small systems of languages such as Hebrew and
Spanish, with five phonological vowels (e.g., /i, e, a, o, u/), can vary in their articulation manner and spectral structure (e.g.,
formants). Such systems with between-vowel contrasts" are "based on a combination of quantity and quality (Becker,
2010:45; see his five-vowel trapeze for Spanish , p. 94, and p. 191 for various languages, including Arabic dialects, Hebrew
and Spanish). Our two CAI dialects reveal that the five long vowels of CAI fit very well into this structure, but more
variation exists in the short vowels as Figure 2 above shows.
ACKNOWLEDGMENT
We cordially thank Rizan Rabi, Sadja Hagala-Asi, Saleem Haj and Najla Kassis for their help in data collecting and
analysis of the MD and GD material.
Rosenhouse et al.
REFERENCES
Abou Haidar, L. (1994) "Norme linguistique et variabilit dialectale: analyse formantique du systme vocalique de la langue arabe,"
Revue de Phontique Applique, 110: 1-15.
Al-Tamimi J. (2002) Variabilit phontique en production et en perception de la parole: Le cas de l'arabe jordano-palestinien.
Universit Lyon: Mmoire du DEA en Sciences du Langage.
Amir, N., Amir, O. and Rosenhouse, J. (submitted) "Colloquial Arabic vowels in Israel: A comparative acoustic study of two dialects"
Amir, N., Tzenker, O., Amir, O., Rosenhouse, J. (2012) "Quantifying Vowel Characteristics in Hebrew and Arabic," paper read at
Afeka Conference for Speech processing.
Becker-Kristal, R. (2010) Acoustic typology of vowel inventories and Dispersion Theory: Insights from a large cross-linguistic corpus,
Ph.D dissertation, UCLA. http://www.linguistics.ucla.edu/faciliti/research/RBecker_diss.pdf
Boersma P. (2001) "Praat, a system for doing phonetics by computer," Glot. International, 5:9/10, 341-345
Bradlow A.R. (1996) "A Perceptual Comparison of the /i//e/ and /u//o/ Contrasts in English and in Spanish: Universal and
Language-Specific Aspects", Phonetica, 53: 5585
Engstrand, Olle (1999) "Swedish", in: Handbook of the International Phonetic Association: A Guide to the usage of the International
Phonetic Alphabet, Cambridge: Cambridge University Press, 140
Escudero, P., Boersma, P., Schurt Rauber, A. and Bion R. A. H. (2009) "A cross-dialect acoustic description of vowels: Brazilian and
European Portuguese, " J. Acoust. Soc. Am. 126(3): 1379-1393.
Ferrari-Disner, S. (1983) Vowel quality, the relation between universal and language-specific factors. UCLA Working Papers in
Phonetics, 58.
Fougeron, C. and Smith, C.L. (1993) "Illustrations of the IPA: French", Journal of the International Phonetic Association, 23 (2): 73
76
Ghazali, S. (1977), Back consonants and backing co-articulation in Arabic, PhD dissertation, University of Texas, Austin
Kaye, A.S. and J. Rosenhouse (1997) "Arabic dialects and Maltese" in: R. Hetzron (ed.) The Semitic Languages, London: Routledge
Publishing House, 263-311.
Labov, W. (2010) Principles of Linguistic Change: Cognitive and Cultural Factors, Wiley-Blackwell: Chichester, UK.
Mangold, M. (2005) Das Aussprachewrterbuch, Mannheim: Duden
MATLAB (2010a) The Mathworks, Natick MS http://www.mathworks.com/products/matlab/
Newman, D.L. and J. Verhoeven (2002) "Frequency analysis of Arabic vowels in connected speech," Antwerp Papers in Linguistics,
100: 77-86.
Rabiner, L. R. and R.W. Schafer (1978) Digital Processing of Speech Signals, Englewood Cliffs, N.J.: Prentice-Hall.
Roach, P. (2004), "British English: Received Pronunciation", Journal of the International Phonetic Association, 34 (2): 239245
Rosenhouse, J. (2002) "Trends of Colloquial Arabic Dialects in Israel," in W. Arnold and H. Bobzin (eds.) "Sprich doch mit deinen
Knechten aramaeisch, wir verstehen es! 60 Beitraege zur Semitistik. Festschrift fr Otto Jastrow zum 60. Geburtstag, Wiesbaden:
Harrassowitz, 599-611.
Saadah, E. (2011) The Production of Arabic vowels by English L2 Learners and Heritage Speakers of Arabic, PhD dissertation, the
University of Illinois, Urbana-Champaign.
Szende, Tams (1994), "Illustrations of the IPA: Hungarian", Journal of the International Phonetic Alphabet, 24 (2): 9194