Journal of Information & Communication Technology

Vol. 3, No. 1, (Spring 2009) 11-20

Phonology for Sindhi Letter-to-Sound Conversion

Javed Ahmed Mahar *

Department of Computer Science,
Shah Abdul Latif University, Khairpur, Pakistan.

Ghulam Qadir Memon *

FEST, HIIT, Hamdard University, Karachi, Pakistan.

ABSTRACT
The Text to Speech (TTS) synthesis technology enables machines to convert
text into audible speech and used throughout the world to enhance the accessibility
of the information. Letter to sound (LTS) conversion is necessary component
of any TTS system and phonological knowledge is essential for LTS conversion.
This study deals with the conversion of Sindhi alphabet letters into their
appropriate sounds. In this paper, phonology of Sindhi language is focused. For
this purpose, some important areas of Sindhi phonology and writing system is
reviewed and presented which can be used for Sindhi letter to sound conversion
and also for the development of rule based Sindhi TTS synthesis system.

INSPEC Classification : C6150, C6170, C6180, C150, C7820.

Keywords : Text to speech, Letter to sound, Phonology, Phoneme, Diphthongs.

1. INTRODUCTION
Sindhi is an Indo-Aryan language of the Indo-European family, related to Hindi, Urdu and
the languages of northwest Indian subcontinent. In Pakistan it is written using a modified
form of the Perso-Arabic script with several additional letters to accommodate Sindhi
implosive, retroflex and nasal sounds. It has many more consonants and vowels than
Arabic. Sindhi occupies a prominent place among the languages of South Asia (Cole,
2005).

Sindhi is an earliest language of sub-continent. According to alphabet some languages like

Urdu and Arabic are the sub-set of Sindhi language unfortunately it has not received the
attention in computational language processing especially in terms of speech synthesis.
In this paper phonology for Sindhi LTS conversion is focused because LTS conversion
module is necessary component of Sindhi TTS system and phonological information is
essential for LTS conversion.

The purpose of TTS synthesis is to convert input text to natural sounding speech as a result
the information will transmit from a machine to a person. TTS systems provide voice
output for all kinds of information such as phone numbers, addresses, navigation information,
*The material presented by the authors does not necessarily portray the viewpoint of the editors
and the management of the Institute of Business and Technology (BIZTEK) or Shah Abdul Latif University,
Khairpur, Pakistan & Hamdard University, Karachi, Pakistan.
*
Javed Ahmed Mahar : mahar.javed@gmail.com
* Ghulam Qadir Memon : gqmemon@hotmail.com

C JICT is published by the Institute of Business and Technology (BIZTEK).

Ibrahim Hydri Road, Korangi Creek, Karachi-75190, Pakistan.
 Javed Ahmed Mahar, Ghulam Qadir Memon

and for reading books (Shah, 2004). TTS is divided into two stages. The first stage takes
text input, processes it and converts it into precise phonetic string to be spoken. The second
stage takes phonetic representation of speech and generates the digital signal.

LTS conversion is always based on some specific language rules. The two main justifications
are conforms the need of LTS component. Firstly, there will always be genuinely new
words in Sindhi language such as: glass, email., table created in the course of time or
adopted in other languages and there are many words which may not be new, but were
ignored when the system was originally built and have now become common enough to
require proper pronunciation such as: bin laden, Obama. Secondly, LTS by rules can be
used in cases where memory is limited.

Phonology is the study of the sound systems of languages. It is concerned with the linguistic
patterning of sounds in human languages. Generally phonology is divided into two branches:
(i) phonetics (ii) phonemics. In phonetics sounds of a language their types, pronunciation
and segmentation are analyzed. The arrangement of phonetic sounds and their linguistically
use is study in phonemics.

2. RELATED WORK
European scholars were the first to attempt a phonological and grammatical analysis of
Sindhi. Their attention was drawn especially to the implosive stops which are unique
characteristics of Sindhi and a few other Indo-Aryan languages. The four implosive stops
in Sindhi were first described by George Stack in 1853. From that time to the present
linguists have, with varying degree of clarity, attempted to describe these sounds. However,
two contemporary linguists Bordie (1958) and Khubchandani (1961) have applied modern
linguistic methods in their analysis and description of Sindhi sounds.

In past, there have been many developments in the Sindhi language particularly in terms
of phonology. The Sindhi phonology, its morphological structure and syntax is discussed
in (Jatoi, 1968). Cole (2005 and 2006) discussed the chart of Sindhi vowel and consonant
sounds with IPA symbols, Sindhi syntax with grammar, morphological sound structure
and Sindhi phonology. Bugio (2001) and Pauline (1981) discussed the consonantal vowel
sounds and its types and present Sindhi Letters and their sounds. TTS synthesis system
for Urdu and Sindhi is designed and developed by Shah et al. (2004) using knowledge
based and hybrid rule based approach. Concatenative synthesis method is selected for this
TTS in which actual snippets of recorded speech is used that were cut from recordings and
stored in voice database. They also presented the phonemes of Sindhi and Urdu. Bird
(1991) investigates Arabic verb morphology, Arabic syllable structure, phonological
constraints and present theory of phonology. Sarfraz et al. (2003) discussed the writing
forms of the Arabic alphabet.

Recently, many research efforts have been put into the field of natural language processing,
including text to speech synthesis systems. The first task in the phonological processing
is to convert the input text into a phonemic string using LTS rules. Hussain (2004) describe
Urdu writing system its phonemic inventory, LTS rules and architecture of NLP for Urdu
TTS. He also discusses Urdu consonantal and vocalic system. Zamirli (2007) proposed
an algorithmic approach for the automatic generation of the stressing in Arabic language
and represents the tonal rules which are employed in the phonetic module. They adapted,
diagrams, generated for the text processing that acting on the size of the sentences to
reading with intonative contours of natural speech. Muhtaseb et al. (2002) defines a set
of Arabic diaphones/sub-syllables for concatenative Arabic TTS synthesis and proposed
Arabic TTS diagram. They discussed speech segmentation rules, classification of Arabic
consonants and types of syllables. Dakkak et al. (2005) introduced a work to incorporate
emotions: anger, joy, sadness, fear and surprise, in an educational Arabic TTS system and
they presents rules for emotion generation.

12 Journal of Information & Communication Technology

Phonology for Sindhi Letter-to-Sound Conversion

3. SINDHI WRITING SYSTEM

The Sindhi writing system, is based on Persian Arabic Script. Sindhi adds its own
modifications in order to symbolize the many sounds not found in Arabic or Persian. For
example, in the Sindhi alphabet, the original Arabic /t/, written , is extended to include
/th/, /T/, and /Th/, written as , , and , respectively. All sounds not found in Arabic.
This was done by taking the basic shape of the letter and adding or rearranging dots. In
this way Sindhi has extended the 28 Arabic characters to 52 so that the sounds unique to
Sindhi may be symbolized.

Because of the rich heredity of Sindhi in its Sanskrit origins, and the later additions of
many Arabic and Persian words, the alphabet contains some sounds which are represented
by more than one letter. Therefore only one sound is associated with any one letter among
them. The letter used is determined by the origin of the words. This makes spelling more
difficult although on the whole Sindhi is very phonetic in its spelling. The following are
the sounds which may be represented by more than one letter:

/t/ , the common letter, and , in words of Arabic origin.

/s/ , common, and , , in words of Arabic origin. , is also found in a few words of
Persian origin.
/z/ , and , common, , and , in words of Arabic origin.
/H/ , common, , found in words of Arabic origin.

Sindhi characters are written from right-to-left. This means that the first letter of a word
appears at the right edge of the word, and the successive letters follow in a leftward
direction. There are 52 distinct letters in the Sindhi alphabet and seven diacritic signs, but
some of these like alifu and small alifu, represent a consonant sound.

The graphic representation of each alphabet of Sindhi, Arabic and Urdu languages has
more than one form depending on its position. Most of the letters have four related forms
(Beginning form BF, Middle form MF, End form and Isolated form). Four forms of Sindhi
letters are described in Table 1. Some letters only connect on one side and are called
"partially connecting" letters. They use just one shape for the initial and medial, and another
shape for final and detached (Sarfraz, 2003) .

Table 1
Four forms of Sindhi letters

3.1 Basic Shape Groups

The 52 letters of Sindhi language are divisible into sixteen basic shape groups. Various
letters may have the same basic shape, but are differentiated from each other within the
group by the use of dots above, within or below the basic shape of the letter.

The four major shape groups are illustrated by these letters:

Letter Group 1
This group contains only /A/. When found at the beginning of a word, the diacritic

Vol. 3, No. 1, (Spring 2009) 13

 Javed Ahmed Mahar, Ghulam Qadir Memon

"madd" will be written over like "aana" /eggs/ . It is not usually found over in the
medial or final position. An important function of is as a "carrier" of other vowels when
a word begins with a vowel. The diacritical marks representing the short vowels must
always be carried by when at the beginning of a word. In other position in the word they
are carried by the relevant consonant symbol.

Letter Group 2
This group contains /b/, /bb/, /bh/, /t/, /th/, /T/, /Th/, /s/, /p/, an partially
/n/, / R /. The letter is an uncommon Arabic consonant that is; it is not frequently
used in Sindhi. The letters /n/ and / R / differ somewhat from others. The forms of
and are more rounded than the others also they drop below the main lines of writing.
The letter also has special forms, initial stands only for the consonant sound /y/, /I/,
/E/, or /ai/ is symbolized by plus . For example, "eiman" /faith/ . Note that the only
difference between /I/ and /E/ sound as symbolized is the inclusion of the diacritic "zer",
with , thus .

Letter Group 3
This group includes /j/, /jj/, /jh/, / N /, /c/, /ch/, /H/, /K/. The letter /H/
occurs only in words of Arabic origin.

Letter Group 4
This group contains /d/, /dh/, /D/, /Dh/, /dd/ and /z/. The letter is an uncommon
Arabic consonant. Thus it is not found frequently in Sindhi.

Letter Group 5
This group contains /r/, /R/, and /z/. The letter is the most common representation
of /z/ in Sindhi. Notice the difference in the shape of the group and that of the group.
Peoples sometimes confuse the two in their writing. The is written with a relatively
closed angle. Also, the drops down below the line of writing and the does not.

Letter Group 6
This group includes /s/ and /S/.

Letter Group 7
This group includes /s/ and /z/. These letters are found only in loan words of Arabic
origin.

Letter Group 8
This group includes /t/ and /z/.

Letter Group 9
This group contains /!/ and /G/. The has no easily assignable phonemic value in
Sindhi. It occurs only in very literary pronunciations of Arabic loan words. Sindhi speakers
usually omit the pronunciation entirely.

Letter Group 10
This group contains /ph/, /f/ and /q/.

Letter Group 11
This group contains only /k/.

Letter Group 12
This group includes /kh/, /g/, /gg/, /gh/, / g /. Before /A/ and /l/, special
initial and medial forms are found. Like "khadho" /food/ ,"bhaggalu" /broken/ etc.
Notice the extra stroke that distinguishes the voiced velar stops from the voiceless .

14 Journal of Information & Communication Technology

Phonology for Sindhi Letter-to-Sound Conversion

Letter Group 13
The only member of this group is /l/.

Letter Group 14
The only member of this group is /m/.

Letter Group 15
This group contains only /v/. When is found at the beginning of a word, it stands for
the consonant sound /v/. When it is used to represent an initial vowel sound, /O/, /U/, or
/ao/ it is found with /A/.For example, "ocito" /sudden/ . In the medial and final
position, may represent either the consonant sounds or any of the vowel sounds. The
only difference in representation between /O/ and /U/ and the diphthong /ao/ is the presence
or absence of the diacritical marks.

Letter Group 16
The only member of this group is /H/, like "hath" /hand/ . This letter also functions
as the symbol for aspiration in /jh/, and /gh/.

/hamzo/ When within a word any syllable ends with a vowel (long or short) and the
following syllable begins with one, the two vowels are separated by /hamzo/ . It serves
the same purpose that a hyphen does in English, that is to separate two syllables. In the
initial and medial forms /hamzo/ must be written over a "carrier" which is the same basic
shape as letter group 2.

4. PHONOLOGICAL ANALYSIS
The phonological systems of Sindhi in most respects resemble that of other Indo-Aryan
languages. Sindhi has a very rich sound inventory. It has 43 distinctive consonant phonemes
and 10 vowels. The phoneme is usually pronounced as an alveolar tap, though occasionally
reminiscent of a trill with two or more contacts. There are three short vowels [a, i, u] and
five long vowels [aa, ii, uu, e, o]. There are also two diphthongs [ai, au], but these are
infrequent and many dialects pronounce these the same as [e, o].

Among the fifty two characters and seven diacritic signs, twenty nine characters are adopted
from Arabic script. Three modified characters adopted from the Persian script: , , .

Twenty modified characters to represent Sindhi sounds:

Retroflex sounds:

Rest: Voiceless Aspirates:

Voiced Aspirates:

Implosive:

Nasal:

4.1 Consonant sounds of Sindhi

There are 50 letters in the Sindhi alphabet that stand for consonant sounds. As some letters

Vol. 3, No. 1, (Spring 2009) 15

 Javed Ahmed Mahar, Ghulam Qadir Memon

represents the same sound discussed above so that total of 43 letters are symbolize the
consonant sounds. Each letter always represents the same sound in Sindhi alphabet
respectively, which makes it very easy to read and sound out new words.

There is no word in Sindhi language in which two or more consonants used in any portion.
For example in English the word "structure" has three consonants in starting position.
Even two consonants are not found in any word of this language. In reference to Sindhi
language word structure, vowels are used mostly after each consonant. Similarly no word
is found ending consonants.

The following are the consonants sounds of Sindhi language (Pauline, 1981).

b ( ): Voiced unaspirated bilabial stop, as in "baby"

bb ( ): Voiced bilabial implosive stop
bh ( ): Voiced aspirated bilabial stop
t ( ): Voiceless unaspirated retroflex stop.
th ( ): Voiceless aspirated retroflex stop.
T ( ): Voiceless unaspirated dental stop; not the English "t" sound, which is alveolar.
Th ( ): Voiceless aspirated dental stop
P ( ): Voiceless unaspirated bilabial stop.
j ( ): Voiced unaspirated palato-alveolar affricate, as in "joy".
jj ( ): Voiced palatal implosive stop.
jh ( ): Voiced aspirated palato-alveolar affricate, as in "judge"
N ( ): Voiced palatal nasal.
c ( ): Voiceless unaspirated palato-alveolar affricate similar to the sound in "cheese" if
pronounced without aspiration.
ch ( ): Voiceless aspirated palato-alveolar affricate, as in "choo-choo".
k ( ): Voiceless velar fricative, similar to the German "ach" and the Scottish "loch"
d ( ): Voiced unaspirated retrofrlx stop.
dh ( ): Voiced aspirated retroflex stop.
dd ( ): Voiced alveolar implosive stop
D ( ): Voiced unaspirated dental stop; not the English "d" sound, which is alveolar
Dh ( ): Voiced aspirated dental
r ( ): Voiced alveolar trill or flap; similar to the Spanish trilled "r."
R ( ): Voiced retroflex flap
z ( ): Voiced alveolar fricative, as in "zebra".
s ( ): Voiceless alveolar fricative, as in "see"
S ( ): Voiceless palato-alveolar fricative, as in "sheep".
! ( ): Glottal stop. This is the "ain" of classical Arabic and in Sindhi it occurs in loan
words from Arabic. It is not emphatically pronounced as a glottal stop in Sindhi.
G ( ): Voiced velar fricative.
f ( ): Voiceless labiodental fricative, as in fish.
ph ( ): Voiceless aspirated bilabial stop
q ( ): Voiceless unaspirated uvular stop. This is the "qui" of classical Arabic and in
Sindhi it occurs in loan words from Arabic. In ordinary Sindhi pronunciation it becomes
/k/.
K ( ): Voiceless unaspirated velar stop, as in "school"
kh ( ): Voiceless aspirated velar stop, as in "kin".
g ( ): Voiced unaspirated velar stop, as in "go"
gg ( ): Voiced velar implosive stop.
gh ( ): Voiced aspirated velar stop
g ( ): Voiced velar nasal. This is one consonant sound, not two as in "bingo". It is similar
to the sounds in "singing".
L ( ): Voiced dental lateral; similar to the sound of "l" in "lean" "feel" when the tongue
is behind the upper front teeth.
M ( ): Voiced bilabial nasal, as in "man"

16 Journal of Information & Communication Technology

Phonology for Sindhi Letter-to-Sound Conversion

N ( ): Voiced dental nasal, some what similar to the English "n", but the tip of the tongue
is behind the upper teeth.
R ( ): Voiced retroflex nasal flap. Curl the tip of the tongue up to the back of the alveolar
ridge, and make as "n" sound as you flap the tongue against the back of the ridge as it
returns to its position behind the lower front teeth.
v ( ): Voiced labio dental fricative, similar to the sound in "vine" but the friction is
weaker.
H ( ): A stream of air passed through the vocal cords. Position of the tongue is determined
by the vowel that follows it.
Y ( ): Voiced palatal approximant, as in "yes".

4.2 Sounds of Sindhi Vowels and Diphthongs

Vowels in any language are more difficult than consonants to describe because they vary
from person to person. Overlooking minor variations, then, it is possible to distinguish
eight vowels and two diphthongs.

Vowels are distinguished from each other by the position of the tongue. The Sindhi vowels
range from the tongue high in the front of the mouth for /I/ to the tongue high in the back
of the mouth for /U/, with the other vowels falling somewhere in between and lower.
Linguists sometimes plot the location of these vowels on a chart. As they are pronounced
in the order of /I/, /i/, /E/, /ai/, /a/, /A/, /ao/, /O/, /u/, and /U/, you will notice that the jaw
is almost closed for /I, that it opens as we move down the chart to /E/; then it opens more
jaw move down the chart to /A/; then it closes again as we move up to /U/. The Sindhi
vowels are described in Table 2 that describes the movement of the jaw and tongue as
Sindhi vowels are formed. The tongue, too, drops progressively lower from a high front
position at the beginning until it reaches a mid-position, then rises progressively at the
back (Jatoi,1968).

Table 2.
The Chart of Sindhi Vowels
Front Central Back
High I U
Lower-high i U
Higher-mid E O
Mid a
Lower-mid ai Ao
Low A

4.3 Sindhi Phonemes

Sindhi may be divided into six major dialects: (1) Siro or Siraiki, spoken in the northern
part of Sindh. (2) vicholi, spoken in the central part of Sindh. (3) Lari, spoken in the
southern part of Sindh. (4) Sasi, spoken in lasbela and the khairthar range on the western
border of Sindh. (5) Thari, spoken in the eastern part of Sindh and the Sindh-Rajastan
border. (6) Kachi, spoken in the Kutch region of Gujrat on the southern border of Sindh
(Cole, 2006). The totals of 520 phonemes of Sindhi language are used for all dialects.
There are 52 letters in Sindhi language, each letter have 10 different sounds by using the
different diacritics and two letters ? and ?. 520x10 makes 520 phonemes. The phonemes
of Sindhi language are described in Table 3.

Vol. 3, No. 1, (Spring 2009) 17

 Javed Ahmed Mahar, Ghulam Qadir Memon

Table 3.
The Phonemes List of Sindhi language

18 Journal of Information & Communication Technology

Phonology for Sindhi Letter-to-Sound Conversion

4.4 Syllables
The concatenation of such letters make syllables and the concatenation of syllables make
words. There is differ of opinion of linguistics for the definition of syllables, all are agreed
that the syllables are exists in words. Every intellectual can differentiate the syllables of
words because his ears are habitual for listening of his language sounds. Ears play important
role for segmentation of words into syllables. Syllabification is not measured in written
form. The concatenation of letters makes syllables only for the sonority of sound.

Syllable division in a word is predictable in Sindhi. Sindhi is primarily an open syllable

language, i.e., syllables mostly end with a vowel or semivowel. Words in Sindhi mostly
have vocalic ending and the occurrence of consonant cluster is also irregular in the language.

A syllable in Sindhi consists of at least one vowel or at most five sounds units, in which
one is a vowel and others are non-vocalic sounds (consonants or semivowels preceding
or following the vowel). Types of syllables in Sindhi are (Jatoi, 1968):

V Vowel
V Long Vowel
C Consonant

. V: Like
. CV: Like
. C V : Like
. C V V: Like
. CCV : Like
. CC V C: Like
. CV C: Like
. CVCC: Like

4.5 Stress
The accent of words is usually change in different languages. Some words start with stress
and few without stress. The languages in which meaningful difference is occur; we say
that these languages have phonemic importance for sound stress. For example in English,
word permit has two syllables (per.mit). If we stress first syllable then the word permit
is considered as a noun means ( ) but if we stress on second syllable then word
permit is considered as a verb means ( ).

In Sindhi, stress has only a limited use of demarcating words and putting emphasis on a
particular word in an utterance. There are three main stresses: word stress, emphatic stress
and drawled stress.

5. CONCLUSION
Letter to sound conversion is a central component of rule based Sindhi TTS synthesis
system. The phonological knowledge is essential for LTS conversion. In this paper, we
have reviewed and formulated the phonology of Sindhi language. This could be support
for building automatic LTS conversion component. For this purpose, we have explained
the details of Sindhi writing system and letters shape groups. Different areas of Sindhi
phonology like: consonant and vowels sounds, Sindhi phonemes, the syllable structure
and importance of stress ness of the language pronunciation are discussed and presented.
This phonological analysis will also support for many other languages like Urdu, Arabic
and Persian.

Vol. 3, No. 1, (Spring 2009) 19

 Javed Ahmed Mahar, Ghulam Qadir Memon

REFERENCE
Bird, S., (1991), "A Logical Approach To Arabic Phonology", Proceedings of the 5th
Conference on European Chapter of the Association for Computational Linguistics,
pp. 89-94.
Bordie, John, (1958), "The Phonology of Sindhi", PhD Theis, University of Texas.
Bugio, M.Q., (2001), "Sociolinguistics of Sindh", LINCOM EUROPA? PhD Thesis.
Cole, Jennifer., (2005), "Sindhi", In Strazny, Philipp(ed) Encyclopedia of Linguistic. New
Yark: Routledge.
Cole, Jennifer, (2006), "The Sindhi Language", In K.Brow(ed) Encyclopedia of Language
and Linguistics, 2nd Edition, v.11: pp. 384-386. Oxford:Flsevier.
Dakkak, O.; Ghneim, N. Abou Zliekha, M.; moubayed, S., (2005), "Emotion Inclusion
in an Arabic Text-to-Speech", 13th European Signal Processing Conference.
Hussain, S., (2004), "Letter-to-Sound Conversion for Urdu Text-to-Speech System",
COLING 2004 Computational Approaches to Arabic Script-based Languages, pp. 74-
79.
Jatoi Ali Nawaz, (1968), "Ilm Lisan Ain Sindhi Zaban", Institute of Sindhalogy, Hyderabad.
Khubchandani, L.M., (1961), "The Phonology and Morphonemics of Sindhi", M.A. Thesis,
University of Pannyslvania.
Muhtaseb, H.; Elshafei1, M; Ghamdi, M., (2002), "Techniques for High Quality Arabic
Speech Synthesis", Informatics and Computer Science: An International Journal,
Vol.140, pp. 255-267.
Pauline. A. Brown, (1981), "Functional Sindhi", Thesis.
Shah A. A.; Ansari, A. W.; Das, L., (2004), "Bi-Lingual Text to Speech Synthesis System
for Urdu and Sindhi", National Conference on Emerging Technology, pp. 126-130.
Sarfraz, M.; Nawaz, S. N.; Khuraidly, A.A, (2003), "Offline Arabic text recognition
system", Proc. of the Int. comference on geometric modeling and graphics.
Zemirli, Z.; Khabet, S; Mosteghanem, M., (2007), "An effective model of streesing in an
Arabic Text to Speech System", IEEE AICCSA, pp. 700-707.

20 Journal of Information & Communication Technology