Scholarly article on topic 'An acoustic investigation of Arabic vowels pronounced by Malay speakers'

An acoustic investigation of Arabic vowels pronounced by Malay speakers Academic research paper on "Computer and information sciences"

CC BY-NC-ND
0
0
Share paper
Keywords
{Arabic / Malay / Vowels / Phonemes / Praat}

Abstract of research paper on Computer and information sciences, author of scientific article — Ali Abd Almisreb, Ahmad Farid Abidin, Nooritawati Md Tahir

Abstract In Malaysia, Arabic language is spoken, and commonly used among the Malays. Malays use Arabic in their daily life, such as during performing worship. Hence, in this paper, some of the Arabic vowels attributes are investigated, analyzed and initial findings are presented based on tokens articulated by Malay speakers as we can consider the spoken Arabic by Malays as one of the Arabic dialects. It is known that in Arabic language there are 28 consonants and 6 main vowels. Firstly, the duration, variability, and overlapping attributes are highlighted based on syllables of Consonant–Vowel with each syllable representing every Arabic consonant with the corresponding vowels. Next, the dispersion of each vowel is examined to be compared with each other along with the variability among vowels that may cause overlapping between vowels in the vowel-space. Results showed that the vowel overlapping occurred between short vowels and their long counterpart vowels. Furthermore, an investigation of the Arabic vowel duration is addressed as well, and duration analysis for all the vowels is discussed, followed by the analysis for each vowel separately. In addition, a comparison between long and short vowels is presented as well as comparison between high and low vowel is carried out.

Academic research paper on topic "An acoustic investigation of Arabic vowels pronounced by Malay speakers"

Journal of King Saud University - Computer and Information Sciences (2016) 28, 148-156

King Saud University

Journal of King Saud University -Computer and Information Sciences

www.ksu.edu.sa www.sciencedirect.com

An acoustic investigation of Arabic vowels c»^

pronounced by Malay speakers

Ali Abd Almisreb, Ahmad Farid Abidin, Nooritawati Md Tahir *

Faculty of Electrical Engineering, Universiti Technologi MARA, 40450 Shah Alam, Selangor, Malaysia

Received 27 February 2015; revised 19 August 2015; accepted 19 August 2015 Available online 31 October 2015

KEYWORDS

Arabic;

Malay;

Vowels;

Phonemes;

Abstract In Malaysia, Arabic language is spoken, and commonly used among the Malays. Malays use Arabic in their daily life, such as during performing worship. Hence, in this paper, some of the Arabic vowels attributes are investigated, analyzed and initial findings are presented based on tokens articulated by Malay speakers as we can consider the spoken Arabic by Malays as one of the Arabic dialects. It is known that in Arabic language there are 28 consonants and 6 main vowels. Firstly, the duration, variability, and overlapping attributes are highlighted based on syllables of Consonant-Vowel with each syllable representing every Arabic consonant with the corresponding vowels. Next, the dispersion of each vowel is examined to be compared with each other along with the variability among vowels that may cause overlapping between vowels in the vowel-space. Results showed that the vowel overlapping occurred between short vowels and their long counterpart vowels. Furthermore, an investigation of the Arabic vowel duration is addressed as well, and duration analysis for all the vowels is discussed, followed by the analysis for each vowel separately. In addition, a comparison between long and short vowels is presented as well as comparison between high and low vowel is carried out.

© 2015 The Authors. Production and hosting by Elsevier B.V. on behalf of King Saud University. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Variability and overlapping of Arabic vowels

While listening to vowels, it seems steady and unchanging due to the fast modification, which may happen in milliseconds in

* Corresponding author. Tel.: +60 179915290. E-mail addresses: alimes96@yahoo.com (A.A. Almisreb), ahmad924@ salam.uitm.edu.my (A.F. Abidin), noori425@salam.uitm.edu.my (N.M. Tahir).

Peer review under responsibility of King Saud University.

Production and hosting by Elsevier

acoustic variables such as the fundamental frequency and formant frequencies. The modifications in acoustic variables may influence the uniqueness and intelligibility of the speaker's voice (Gordon, 2012). Variability in the production of vowels may include stress, context, speaking rate and formant frequencies. Three factors could influence the formant frequencies (Nicolaidis, 2003; Seung-Jae and Lindblom, 1994) namely:

(1) Duration of the vowel.

(2) Contextual environment.

(3) Articulatory effort.

There are two phonetic variables to describe the vowels specifically the quality and the quantity of the vowels. The

http://dx.doi.org/10.1016/j.jksuci.2015.08.003

1319-1578 © 2015 The Authors. Production and hosting by Elsevier B.V. on behalf of King Saud University. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

vowel quality describes the vowel's articulation place, the tongue position inside the vocal tract, stricture size, the lips shape and the status of the vowel, whether nasalized or not, while the vowel quantity describes the vowel duration (Saadah, 2011). In addition, the vowel duration could be affected by the speech rate, for instance the vowel duration becomes shorter if the speech rate increases. Moreover, the speakers tend to produce scattered vowels if they spoke slowly and the vowels will be centralized in the vowel space if the subjects spoke faster (Souza and De Mora, 2014).

As stated earlier, there are six vowels in Modern Standard Arabic (MSA), and it can be divided into two categories. Firstly, are the short vowels, which include /a/, /i/ and /u/, and next are the long vowels which is comprised of /a:/, /i:/ and /u:/. Several researches have been carried out about Arabic vowels. A study by Saadah (2011) has investigated the production of Arabic vowels by English second language (L2) learners and heritage speakers of Arabic. Another study by Thesieres (2001) has addressed the vowels in Lebanese Arabic and the United Arab Emirates. Results of Lebanese Arabic vowels were compared with the results of Iraqi Arabic vowels based on a study conducted by Al-Ani (1970). Meanwhile, the experiments on Emirates Arabic vowels were compared to the Lebanese and Iraqi vowel experiments.

An investigation on the acoustic attributes of the Palestinian Arabic vowels was done by Saadah (2011). This research was based on tokens that have been articulated by six native Palestinian Arabic speakers with a total number of vowel tokens as 1368. His experimental results showed that F1 for short /i/ and short /u/ has a higher frequency than its counterpart vowel, and this refers to the high long vowels, which were produced with higher tongue position. However, the short low vowel /a/ has a lower F1 frequency compared to its long counterpart vowel. Meanwhile, for F2 frequencies, the speakers were more likely to articulate the short /i/ with a lower value than the long /i/ in contrast with the short /u/. The Palestinian vowels have lower F1 and F2 values compared to Iraqi vowels (Al-Ani, 1970) and Tunisian vowels (Belkaid, 1984). Another research on vowels in the Palestinian Arabic was conducted by Adam (2014) which aimed to study the variation in the vowel durations in two cases: normal speakers and speakers with Broca's aphasia. The study claimed that the vowel durations were longer for the speakers with Broca's aphasia compared to normal speakers. Researchers have also focused to study the vowels in other Arabic dialects. For example, Saudi, Sudanese and Egyptian Arabic vowels have been addressed by Alghamdi (1998) and its aim is to decide whether vowels in MSA are realized in the same way if spoken by individuals related to different dialects. The researcher found that the short vowels were likely to be centralized more than the long vowels. Another research on vowels in eight Arabic dialects is conducted by Haidar (1994). The dialects include Lebanese, Syrian, Qatari, Tunisian, Emirati, Jordanian, Saudi, and Sudanese. The researcher used monosyllabic words in her experiments. This study has shown a significant difference in the formant values among all eight dialects. Another study of the vowels in the Libyan Arabic is addressed by Ahmed (2008). The aim of this study is to provide acoustic and auditory descriptions about vowels in Libyan Arabic in order to compare it with vowel's attributes of other Arabic dialects. The use of monosyllabic words was recorded among 20 native Libyan Arabic native speakers. His results showed that the

long and short vowels were significantly varied in both quantity and quality. In case of the short vowels, it was likely to be more centralized compared to other results reported by other studies. Formant based analysis of spoken Arabic vowels is also studied by Alotaibi and Husain (2009). The first two formants were considered in this study, in addition to the differences and similarities between vowels. All the carrier words were formed using Consonant-Vowel-Consonant style (CVC).

In addition, an analysis study of the formant frequencies of the Arabic vowels is achieved by Newman and Verhoeven (2002). This research was based on Quranic recitation tokens, which consists of 30 min of Qur'an recitation. The recorded token contains 400 vocalic observations, which cover all the Arabic vowels. Moreover, along with the Quranic recitation tokens, the researchers acoustically analyzed the same vowels depending on recorded tokens taken from colloquial Egyptian Arabic.

2. Duration of Arabic vowels

Every speech sound has its physical and perceptual properties. The perceptual values of any speech sound can be linked to the physical value measured. Duration can be defined as the physical property that represents the measured duration of a speech sound from the articulatory and acoustic points perspective (Hassan, 1981). From another point of view, the duration of a speech sound can also be in the representation of time dimension of an acoustical signal (Lehiste, 1970). On the other hand, length is defined as the perceptual attribute that leads to the perception of a speech sound. Several researches have done in depth investigation on the durational and articulatory parameters in vowel articulation. Some significant information based on the findings include the vowel duration, such as the ability of the listener to perceive vowels and the production mechanism of the vowel or even demonstrate the articulatory movements degree that are required for producing a particular vowel (House, 1961). Moreover, the acoustic studies have claimed that the vowel duration was beneficial for vowel identification and the intelligibility of speech (Ferguson and Kewley-port, 2007; Mok, 2011). The acoustic investigations have reported that the low vowels are longer than the high vowels, while vowels produced within closed syllables are shorter than vowels produced within opened syllables. In addition, vowels followed by voiced consonant phonemes were found to be longer compared with vowels followed by voiceless consonant phonemes. Alternatively, vowels before stop consonant phonemes are shorter than vowels followed by fricative consonant phonemes (Ladefoged, 2006) and because vowels in Arabic are a concern, a study on Arabic vowels was carried out by Alghamdi (1998). The vowels were pronounced by speakers representing three Arabic dialects: Saudi, Sudanese and Egyptian. The experimental results have shown that the main difference between the three dialects was in the first formant frequencies, but in terms of duration, the vowels did not show any significant difference from one another. In addition, it was found that the behavior of the short vowels was less than half of the long vowels. Another study on vowels in Arabic dialects has been addressed by Khattab and Al-tamimi (2007) for Lebanese Arabic. This study reported that there is no significant difference between the durational results

for males and females. Ammani-Jordanian Arabic has been studied by Zawaydeh (1997) and it has shown that the frequencies for the F1 formant in the low vowels in the uvu-larised environment are higher than F1 frequencies in the plain environment.

3. Methods

This section discussed the proposed method. A group of eight Malay speakers that comprised of 6 males and 2 females, ages ranged between 18 through 38 years is the database used in this research. The speakers are from different regions in Malaysia which are; Selangor, Melaka, Perlis, Johor Baharu, Perak, Pahang, Sabah, and Sarawak. In addition, all the subjects are students. Each subject is required to read and record all tokens in one session. The tokens have been pre-prepared in advance, and it consists of Consonant-Vowel (CV) syllables. Therefore, in total the number of syllables articulated by each speaker is 168 (28 consonants x 6 vowels) and the total number of recorded vowels is 1344 (168 tokens x 8 speakers).

The recording process is accomplished using SAMSON C03U USB multi-pattern condenser microphone. This microphone has the ability to record high-quality voice even in a noisy environment due to its built-in switchable high-pass filter and 10 dB pad. The chosen sampling rate for all the recording phonemes is 16,000 Hz, and 32 bits for mono channel. Audacity 2.0.3 software is used as the recording platform.

Table 1 presents the carrier vowels among all the Arabic alphabets and its International Phonetic Alphabets (IPA) representation.

4. Results based variability and overlapping

In this section, analysis of the vowels was conducted using Praat software (Boersma and Weenink, 2014).

4.1. Between-vowel category variability

Firstly, it was found that there exist variations between vowel categories. In order to compare the within-category variability between all categories, the dispersion range of each vowel category for both formants is calculated in Hz. Then, the dispersion for all vowel categories is presented using the bar graph as depicted in Fig. 1.

As observed in Fig. 1, the central mid vowel /a/ appears to be the most dispersed vowel on the front back dimension with F2 in the range of 2911.42 Hz. On the other hand, the least dispersed vowel seems to be the high front vowel /i:/, with F2 range at 510.176 Hz. The least dispersed vowel is F3 since it has the least range difference of 320.376 over all other vowels.

The fact that front vowels are less dispersed than back vowels is explained by Beckman et al. (1995) that suggested the articulatory configuration of /i/ is easier to obtain than that of /u/, resulting in less variability in the formants. High front vowels can be pronounced more precisely by stiffening the genioglossus muscle and sustaining the tongue laterally against the dental edge. By contrast, the articulation of the back vowels /u/ are not able to be obtained in a similar manner, leading to the less precise control of tongue height, which results in more dispersion in the production of this vowel.

Table 1 Carrier vowels.

Short fatha Long fatha Short dummah Long dummah Short kasrah Long kasrah

Vowel IPA Vowel IPA Vowel IPA Vowel IPA Vowel IPA Vowel IPA

i aa i! aaa: i au ji auu: i ai ¡i aii:

y ba U baa: y bu js buu: y bi vS bii:

ta ii taa: ¿J tu tuu: ti v5 tii:

à 0a ii 0aa: à 0u j3 0uu: à 0i v? 0ii:

É d3a u d3aa: é d3u Jt d3uu: -t d3i vt d3ii:

t ha u haa: t hu j»- huu: hi v hii:

é xa n xaa: é xu j4 xuu: -4 xi v4 xii:

da is daa: du j3 duu: 3 di ¡3 dii:

j öa ii öaa: j öu jj öuu: j öi ¡j öii:

j ra ij raa: j ru Jj ruu: j ri rii:

j za ij zaa: j zu jj zuu: j zi ¡j zii:

i> sa u, saa: t su ji suu: -i si sii:

ii Ja ii Jaa: l Ju Juu: -4 Ji v5 Jii:

d-= s% i^ s%a: L)-= s% sfuu: o? sfi v^1 sfii:

LÖ d% li d%a: O d% j^ dfuu: o dfi dfii:

1 t% ii t%a: i t% ji tfuu: 1 tfi vi tfii:

i Öf a ii ö%a: i Öf u ji öf uu: i öf i vi Öf ii:

Ça i^ Çaa: Çu ji Çuu: Çi Çii:

iL Ka ii Kaa: -i Ku Jf Kuu: _f Ki v1 Kii:

L fa ii faa: L fu J? fuu: LL fi V? fii:

(J qa ii qaa: (J qu JÎ quu: J qi v^ qii:

¿1 ka ii kaa: ¿1 ku js kuu: ki VS kii:

J la V laa: J lu jj luu: J li vJ lii:

ma u maa: mu ji muu: mi vi mii:

j na ii naa: j nu ji nuu: -i ni vi nii:

i ha ii haa: i hu J* huu: i hi v* hii:

J wa ij waa: J wu jj wuu: J wi ¡J wii:

i ja ij jaa: ¡ ju JJ juu: -J ji vJ jii:

Figure 1 Vowel category dispersion (Range).

However, although range might give some idea about the dispersion of vowels, especially with regard to minimum and maximum points, it might not be the optimal measurement since there may exist some outliers that affect the measured dispersion in spite of the fact that they might be few in number. Another way to measure dispersion via computation of standard deviation is shown in Fig. 2 below.

As compared to Figs. 1 and 2 depicted similar plot patterns and hence minimal differences in the overall dispersion pattern. Here vowel /a/ is still the most dispersed vowel specifically F2, followed by the short vowel /u/ that is less dispersed. However, the vowel /a/ is still the most dispersed vowel as far as Fl is concerned while the vowel /i/ is the most dispersed vowel if F5 is considered. The least dispersion is shown by the long vowel /i:/. From these two plots, it can be summarized that long vowels are less dispersed than short vowels and front vowels are less dispersed than back vowels.

The tendency for high front vowels to be less dispersed than other vowels might be explained by the fact that most of the consonants preceding and/or following all vowels in the data material are coronals, which have an anterior place of articulation, which is similar to that used in the articulation of the high front vowels. The tongue does not need to move over a long distance to/from the consonant in order to produce the vowel. This leads to less influence on front rather than back vowels and makes them less dispersed. Moreover, low vowels

require jaw lowering, which requires more time for the tongue to move to/from the articulatory position of a low vowel, which makes this vowel more liable to consonantal effects.

4.2. Vowel quality overlap

Here, overlapping of vowels will be elaborated. Some vowels were found to be overlapped and this might be partially attributed to the high degree of variability these vowels exhibit. For example, the acoustic space for the long and short dummah vowels overlapped one another. Similarly, the long and short kasrah vowels overlapped one another too. Fig. 3 showed the overlapping between /u/ and /u:/ while Fig. 4 depicted the overlapping between /i/ and /i:/.

Acoustic overlap in the vowel space is common. For example, Peterson and Barney (1952) found a considerable overlap in the production of American vowels by native speakers. This overlap was attributed to several factors namely the differences in the vowel tract size and the context in which these vowels are produced (Ryalls, 1996). When there is variation in the vowel produced by the same speaker, listeners use structural estimation (Nusbaum and Morin, 1992). That is, they rely on the different cues found in the vowel produced to identify it. For example, listeners benefit from F0 and F3 in addition to F1 and F2 in order to recognize the vowel. However, when different vowels are produced by the same speaker sound

Figure 2 Vowel category dispersion (SD).

Figure 3 Overlapping between /u/ and /u:/.

similar, listeners rely on a contextual tuning mechanism (Nusbaum and Morin, 1992). In this kind of normalization, listeners try to benefit from contextual information external to the vowel. Information relating to other sounds and utterances found near that vowel is used as an additional cue found in higher levels including lexical, grammatical, and semantic

5. Vowel duration measurements

As claimed by Flege and Port (1981), measuring of vowel duration is from the onset of energy in F1 to the offset of energy in F1 and F2, which with vowel boundaries can be determined from the start to the end. Fig. 5 illustrates a sample of vowel waveform and its corresponding spectrogram. This figure demonstrates the location of the vowel and the area between the two lines representing the duration of the vowel.

5.1. Reliability of the vowel duration measurements

Reliability is mainly concerned with the consistency of measurement done, that is, whether the measurements taken are consistent and repeatable (Bryman, 2001). With regard to the production results and in order to check the reliability of the formant measurements, suggestions made by Ladefoged (2003) are followed.

Foremost, upon completion of all measurements, repetitions of measurement are done too. Due to the large number of tokens, it required about one month to complete all

Figure 4 Overlapping between /i/ and /i:/.

measurements and one more month to repeat them. The two repetitions are then compared, and if there is a significant divergence between the first and the second measurements of a certain formant, a third measurement is made to ensure which is more precise among the two.

Since all tokens were recorded twice, the second step is to ensure reliability to compare the formants of the two repeated tokens. If, there is a difference of more than 50 Hz between the two repetitions of the same formant, they were checked once more to ensure that there was no error in the measurements.

The next step is ensuring the reliability of the measurements that is by plotting all tokens uttered by a specific speaker for a certain vowel on the formant plane. This process would highlight any outliers in the production of each speaker's vowel. When an outlier is found, the formant measurement of that outlier is taken again. If no errors or outliers are found, the data are accepted as showing intra-speaker variability. The same procedure is done for duration measurements.

5.2. Duration analysis results

In order to present and discuss the results of the duration analysis, the overall duration of all vowels is presented first before looking closely at the duration patterns of each vowel separately.

As examined in Table 2, it was revealed that Arabic vowels could be divided into two groups, that is as far as duration is concerned known as short and long vowels. The short vowels are three specifically /a/, /u/ and /i/. On the other hand, the long vowels are the other three that are /a:/, /u:/ and /i:/.

The mean duration of the three short vowels is 0.376 s while that for the three long vowels is 0.689 s. Table 2 also showed the directional similarity within these two groups. The high front /i:/ has the shortest duration among long vowels while the low front /a/ has the shortest duration among the short vowels.

High vowels being shorter than low vowels are attributed to the extra time needed for lowering the jaw when low vowels are produced (Lehiste, 1970; Lindblom, 1967). As presented in Table 2, Malay speakers tended to pronounce long vowels longer than short vowels and high short vowels longer than low short vowels.

Next, Fig. 6 shows that there is a considerable variation in durational patterns between short and long vowels. However, the distinctive difference in duration between short vowels and long ones is still obvious. It is observed that if the duration of a short vowel is high, the duration of a long counterpart produced by the same speaker is still significantly higher in order to preserve the distinction in duration between short and long vowels.

Table 3 shows all durations of short fatha vowel with the Arabic phonemes. The average mean of these durations is 0.364 s. The longest duration is 0.547 s for the short fatha with phoneme /z (j)/, which is a voiced phoneme. On the other hand, the shortest duration is 0.251 s for the short fatha with the phoneme /5 (i)/, which is also a voiced phoneme. As observed in Table 3, for short fatha vowels both shortest and longest durations belong to voiced phonemes.

Table 4 shows all durations of the long fatha vowel with the Arabic phonemes. The average mean of these durations is 0.690 s. The longest duration is 0.939 s for the long fatha with

Time (s)

Figure 5 Waveform and spectrogram illustrate the vowel place and duration measurements.

Table 2 Statistical measurements of vowel durations in

seconds.

Vowels /a/ /a:/ /u/ /u:/ /i/ /i:/

Max 0.547 0.939 0.521 0.901 0.558 0.882

Min 0.251 0.527 0.231 0.524 0.231 0.505

Range 0.296 0.412 0.29 0.377 0.327 0.377

Mean 0.364 0.690 0.386 0.691 0.378 0.688

Ratio 0.527 0.558 0.550

SD 0.079 0.099 0.096 0.100 0.084 0.096

phoneme /J 0-^)/, which is a voiceless phoneme. In contrast, the shortest duration is 0.527 s for the long fatha with the phoneme /t (5)/ which is also voiceless phoneme. As noticed, for long fatha vowels both shortest and longest durations belong to voiceless phonemes.

Table 5 shows all durations of the short dummah vowel with Arabic phonemes. The average mean of these durations is

0.386 s. The longest duration is 0.521 s for the short dummah with phoneme /5! (—)/ which is a voiced phoneme. It is observed that the shortest duration is 0.231 s, for the short dummah with the phoneme /5 (i)/ also as the voiced phoneme. It is also seen that for short dummah vowels, both shortest and longest durations belong to voiced phonemes.

Table 6 shows all the durations of long dummah vowel with the Arabic phonemes. The average mean of these durations is 0.691 s. The longest duration is 0.901 s for the long dummah that is the phoneme /s! (>«)/ which is a voiceless phoneme while the shortest duration is 0.524 s for the long dummah with the phoneme /t (jS)/ which is also a voiceless phoneme. This showed that for long dummah vowels both shortest and longest durations belong to voiceless phonemes.

Table 7 shows all durations of the short kasrah vowel with Arabic phonemes. The average mean of these durations is 0.378 s. The longest duration is 0.558 s for the short kasrah with phoneme /x (-4)/, which is a voiceless phoneme, while the shortest duration is 0.231 s for the short kasrah with the

Figure 6 Long and short vowel distribution.

Table 3 Short fatha durations.

Short fatha Duration Short fatha Duration

i aa 0.503 D d% 0.285

y ba 0.389 J= t% 0.277

ta 0.393 i Öf a 0.365

î 0a 0.279 Ça 0.390

i! d3a 0.358 i Ba 0.311

ii ha 0.279 fa 0.264

ii xa 0.304 6 qa 0.322

da 0.297 cil ka 0.304

j öa 0.251 J la 0.420

J ra 0.380 é ma 0.390

j za 0.547 ù na 0.420

i> sa 0.432 s ha 0.286

uU Ja 0.467 j wa 0.404

d-= s% 0.501 s ja 0.382

Table 5 Short dummah durations.

Short dummah Duration Short dummah Duration

i au 0.445 d% 0.448

y bu 0.262 _i t% 0.263

i tu 0.285 i Öf u 0.521

î 0u 0.289 t Çu 0.424

i d3u 0.319 _f Bu 0.449

i hu 0.236 lj fu 0.242

i xu 0.516 6 qu 0.369

du 0.251 l ku 0.395

j öu 0.231 J lu 0.456

J ru 0.395 é mu 0.459

j zu 0.514 ù nu 0.466

su 0.500 s hu 0.419

u Ju 0.504 J wu 0.347

s% 0.430 s ju 0.381

phoneme /h (;)/ which is also a voiceless phoneme. As observed in Table 7, for short kasrah vowels both shortest and longest durations belong to voiceless phonemes.

Table 8 shows all durations of the long kasrah vowel with Arabic phonemes. The average mean of these durations is 0.688 s. The longest duration is 0.882 s for the long kasrah with the phoneme /J (^i)/, which is a voiceless phoneme while the shortest duration is 0.505 s for the long kasrah with the phoneme /9 (Ji)/, which is also a voiceless phoneme. As tabulated in Table 8, for long kasrah vowels both shortest and longest durations belong to voiceless phonemes.

5.3. Long versus short vowels duration

In this section, short vowels and their long counterparts will be compared. That is, the vowels /a, u, i/ will be compared with / a:, u:, i:/. First, some descriptive statistics of the duration are introduced earlier in Table 2.

As tabulated in Table 2, long vowel duration owned more than twice as compared to their short counterparts. The average mean for short vowels is 0.376 s and that for long vowels is 0.689 s, with a ratio of 0.545. Duration seems to represent a more robust distinguishing factor between short and long vowels. Table 2 also shows that the shortest duration is exhibited

Table 6 Long dummah durations.

Long dummah Duration Long dummah Duration

ji auu: 0.665 dfuu: 0.746

Ji buu: 0.601 ji tfuu: 0.640

tuu: 0.524 ji öf uu: 0.841

0uu: 0.592 ji Çuu: 0.601

d3uu: 0.587 jf Buu: 0.799

j*- huu: 0.665 ji fuu: 0.565

xuu: 0.696 jî quu: 0.591

j3 duu: 0.610 ji kuu: 0.686

jj öuu: 0.605 ji luu: 0.745

jJ ruu: 0.764 ji muu: 0.845

jj zuu: 0.686 ji nuu: 0.696

ji suu: 0.875 ji huu: 0.638

ji Juu: 0.785 jj wuu: 0.765

sfuu: 0.901 ji juu: 0.645

Table 7 Short kasrah durations.

Short kasrah Duration Short kasrah Duration

i ai 0.466 dfi 0.391

y bi 0.336 i й 0.358

i ti 0.307 i öf i 0.278

î 0i 0.333 Çi 0.348

d3i 0.368 Bi 0.428

hi 0.454 fi 0.254

xi 0.558 J qi 0.352

3 di 0.299 4 ki 0.298

j öi 0.350 J li 0.411

J ri 0.293 _i mi 0.451

jj zi 0.301 4 ni 0.426

si 0.527 s hi 0.231

_i Ji 0.512 j wi 0.364

sfi 0.476 -i ji 0.431

by the high front long vowel and low front short vowels. For instance, the difference between the high long vowel /i:/ and the nearest duration, that is found in the low front vowel /a:/ is 0.002 s. On the other hand, the durational difference between the low short vowel /a/ and the nearest duration in the group of short vowels, which is found in /i/, is 0.014 s.

Table 4 Long fatha durations.

Long fatha Duration Long fatha Duration

li aaa: 0.640 Li dfaa: 0.683

13 baa: 0.654 li t%a: 0.642

taa: 0.527 li öf aa: 0.751

IÎ 0aa: 0.547 le Çaa: 0.560

U d3aa: 0.538 li Baa: 0.814

U haa: 0.753 li faa: 0.605

U xaa: 0.788 lî qaa: 0.636

13 daa: 0.599 l£ kaa: 0.666

lj öaa: 0.723 V laa: 0.639

IJ raa: 0.770 li maa: 0.669

lj zaa: 0.729 li naa: 0.763

li saa: 0.810 li haa: 0.618

li Jaa: 0.939 lj waa: 0.690

Li s%a: 0.832 li jaa: 0.744

Table 8 Long dummah durations.

Long kasrah Duration Long kasrah Duration

aii: 0.688 dfii: 0.751

vi bii: 0.670 tfii: 0.702

v tii: 0.572 Öf ii: 0.808

0ii: 0.505 îii: 0.626

vt d3ii: 0.710 v? Kii: 0.736

V hii: 0.575 fii: 0.614

xii: 0.761 vi qii: 0.615

dii: 0.799 kii: 0.686

& öii: 0.583 J lii: 0.636

rii: 0.602 mii: 0.846

zii: 0.649 nii: 0.728

sii: 0.873 hii: 0.633

Iii: 0.882 wii: 0.603

sfii: 0.749 jii: 0.665

However, the high back long vowels and low long vowels do not show much difference in their duration.

5.4. High versus low vowel duration

Lowering the jaw is known to have a positive effect on vowel duration (Klatt, 1976; Lindblom, 1967). Therefore, it is expected that low vowels would be longer than high vowels due to the amount of jaw lowering required by the production of low vowels. Conversely, Malay speakers tend to articulate low short vowels shorter than high long vowels whereby /a/

shorter than /u/ and /i/ by 0.027 and 0.014 s respectively. While in the case of long vowels, the low long vowel /a:/ is longer than the high front long vowel /i:/ but the difference is only 0.002 s. In contrast, the low long vowel /a:/ is shorter than the high back long vowel /u:/ and the difference is 0.001 s. The finding that the low short vowels are unexpectedly shorter than its high counterpart can be partially attributed to the fact that the Malay speakers do not lower their jaw during the pronunciation of the low short vowel.

Eventually, comparisons in terms of the number of participants, as presented in Table 9, and in terms of duration, as illustrated in Table 10, are accomplished between several studies of Arabic vowels and the current study. Several dialects that have been investigated in a single study such as Algeria, Tunisia, Libya, Egypt, Jordan, and Iraq are studied by Ghazeli (1979) and Qatar, Lebanon, Saudi Arabia, Tunisia, Syria, Sudan, United Arab Emirates, and Jordan by Haidar (1994). Moreover, different researchers addressed the same dialect such as Jordanian dialects which is addressed by Al-Ani (1970), Ghazeli (1979), Haidar (1994). Table 9 includes several studies that were carried out to explore Arabic vowel properties. It involves the conducted dialect and number of participants in each study.

Another comparison is conducted based on the duration of various Arabic dialects and the current study. Table 10 shows that Malay speakers articulated Arabic vowels slower than other native speakers. The study addressed by Al-Ani (1970) showed closer vowel length to the vowels articulated by Malay speakers.

Table 9 A comparison between several studies that handled the Arabic vowel topic in term of dialects and number of participants.

Research by Dialect Participants

Al-Ani (1970) Ghazeli (1979) Haidar (1994) Alghamdi (1998) Newman and Verhoeven (2002) Ahmed (2008) Alotaibi and Husain (2009) Saadah (2011) This study Iraqi Jordanian Algeria, Tunisia, Libya, Egypt, Jordan, and Iraq Qatar, Lebanon, Saudi Arabia, Tunisia, Syria, Sudan, United Arab Emirates, and Jordan Saudi Arabia, Sudan, and Egypt Quran vowels Cairene Libya Saudi Arabia Palestine Arabic by Malay speakers 8 2 12 8 (1 participants per dialect) 15 (5 participants per dialect) 1 1 20 10 (9 males and 1 child) 6 8

Table 10 Vowel duration (in seconds) of several Arabic dialects. Study Dialect Vowels

i i: a a: u u:

Al-Ani (1970) Iraqi 0.300 0.600 0.300 0.600 0.300 0.600

Mitleb (1984) Jordanian 0.076 0.116 0.090 0.145 0.083 0.124

Hussain (1985) Gulf Arabic 0.085 0.155 0.106 0.190 0.093 0.165

Alghamdi (1998) Saudi 0.111 0.248 0.133 0.311 0.114 0.137

Sudanese 0.117 0.275 0.128 0.295 0.116 0.304

Egyptian 0.098 0.255 0.122 0.316 0.110 0.253

Ahmed (2008) Libyan 0.054 0.138 0.063 0.150 0.064 0.148

Saadah (2011) Palestine 0.084 0.219 0.097 0.247 0.090 0.226

This study Arabic by Malay 0.378 0.688 0.364 0.690 0.386 0.691

6. Conclusion

As a conclusion, the investigation of Arabic vowel properties is conducted with Malay individuals articulated the carrier tokens. Duration, variability, and overlapping attributes of the Arabic vowels among the speakers have been discussed. The presented attributes have been addressed according to articulated vowels covering all the Arabic phonemes, which lead to an overview on how the pronunciation of the vowels has been done. This study will be able to assist researchers to suggest suitable manners for articulating Arabic vowels and a deeper understanding of the pronunciation process. Additionally, it will also assist to investigate the differences and similarities of pronunciation between native speakers and non-native speakers of the Arabic language.

References

Adam, H., 2014. Acoustical analysis of vowel duration in Palestinian Arabic speaking aphasics. Am. J. Psychiatry Neurosci. 2, 13. http:// dx.doi.org/10.11648/j.ajpn.20140201.13. Ahmed, A., 2008. Production and Perception of Libyan Arabic

Vowels. Newcastle University, pp. 153-167. Al-Ani, S.H., 1970. Arabic Phonology: An Acoustical and Physiological Investigation. Mouton, Paris. Alghamdi, M., 1998. A spectrographic analysis of Arabic vowels - a

cross-dialect study. J. King Saud Univ. 10, 3-24. Alotaibi, Y.A., Husain, A., 2009. Formant based analysis of spoken Arabic vowels. In: Lecture Notes in Computer Science. Springer, pp. 162-169.

Beckman, E., Jong, D., Krishnamurthy, K., Ahalt, C., Collins, M.J., 1995. Variability in the production of quantal vowels revisited. J. Acoust. Soc. Am. 97, 471-490. Belkaid, Y., 1984. Arabic vowels, modern literature, spectrographic analysis. Trav. l'Institut Phonétique Strasbg. Strasbg. 16, 217-240. Boersma, P., Weenink, D., 2014. Praat: doing phonetics by computer [Computer program]. Version 5.4, retrieved 4 October 2014 from < http://www.praat.org/ >. Bryman, A., 2001. Social Research Methods, Second ed. Oxford University Press.

Ferguson, S.H., Kewley-port, D., 2007. Talker differences in clear and conversational speech: acoustic characteristics of vowels. J. Speech Lang. Hear. Res. 50, 1241-1256. Flege, J., Port, R., 1981. Cross-language phonetic interference: Arabic

to English. Lang. Speech 24, 25-146. Ghazeli, S., 1979. Du statut des voyelles en arabe. Analyses-Theories,

Etudes Arabes 2-3, 199-219. Gordon, A., 2012. The Effect of Speaking Style on Variability of Vowel Production for Native Monolingual English Speakers. University of South Florida, pp. 2-19. Haidar, L.A., 1994. Norme linguistique et variabilite dialectale: analyse formantique du systeme vocalique de la langue arabe. Rev. Phon. Appl. 110, 1-15.

Hassan, Z.M., 1981. An experimental study of vowel duration in Iraqi spoken Arabic. Zhurnal Eksp. i Teor. Fiz. University of Leeds. 163-246.

House, A.S., 1961. On vowel duration in English. J. Acoust. Soc. Am. 33, 1174-1178.

Hussain, A.A.A., 1985. An experimental investigation of some aspects of the sound system of the Gulf Arabic dialect with special reference to duration (Ph.D. thesis). University of Essex.

Khattab, G., Al-tamimi, J., 2007. Durational cues for gemination in Lebanese Arabic. 1-12.

Klatt, D.H., 1976. The linguistic uses of segmental duration in English: acoustic and perceptual evidence. J. Acoust. Soc. Am. 59, 12081221.

Ladefoged, P., 2003. Phonetic data analysis: an introduction to field work and instrumental techniques. Blackwell Pub, USA.

Ladefoged, P., 2006. A Course in Phonetics, fifth ed. Wadsworth, Boston.

Lehiste, I., 1970. Suprasegmentals. MIT Press, Cambridge.

Lindblom, B., 1967. Vowel duration and a model of lip mandible coordination. STL-QPSR 8, 001-029.

Mitleb, F.M., 1984. Voicing effect on vowel duration is not an absolute universal. J. Phonetics 12, 23-27.

Mok, P.P.K., 2011. Effects of vowel duration and vowel quality on vowel-to-vowel coarticulation. Lang. Speech 54, 527-545. http:// dx.doi.org/10.1177/0023830911404961.

Newman, D.L., Verhoeven, J., 2002. Frequency Analysis of Arabic Vowels in Connected Speech. Antwerp Pap. Linguist. 100, 65-75.

Nicolaidis, K., 2003. Acoustic variability of vowels in Greek spontaneous speech. In: Proc. 15th Int. Congr. Phonetic Sci. pp. 32213224.

Nusbaum, H.C., Morin, T.M., 1992. Paying attention to differences among talkers. In: Tohkura, Y., Vatikiotis-Bateson, E., Sagisaka, Y. (Eds.), Speech Perception, Production and Linguistic Structure. Burke IOS Press, pp. 113-123.

Peterson, G.E., Barney, H.L., 1952. Control methods used in a study of the vowels. J. Acoust. Soc. Am. 24 (2), 175-184.

Ryalls, J., 1996. A Basic Introduction to Speech Perception. Singular Publishing Group Inc, London.

Saadah, E., 2011. The Production of Arabic Vowels by English L2 Learners and Heritage Speakers of Arabic. University of Illinois, pp. 72-92.

Seung-Jae, M., Lindblom, B., 1994. Interaction between duration, context, and speaking style in English stressed vowels. J. Acoust. Soc. Am. 96, 40-55.

Souza, H.K. De Mora, J.C., 2014. Effects of speech rate on the intelligibility of non-native vowels. In: Proc. Int. Symp. Acquis. Second Lang. Speech Concordia Work. Pap. Appl. Linguist. pp. 116.

Thesieres, H., 2001. An Articulatory Phonological Analysis of Vowel Phonology in Spoken MSA. Rice University, pp. 13-65.

Zawaydeh, B.A., 1997. An acoustic analysis of uvularization spread in Ammani-Jordanian Arabic. Stud. Ling. Sci. 27 (1), 185-200.