Context Sensitive Shape-Substitution in

Nastaliq Writing System: Analysis and

Formulation
Aamir Wali Sarmad Hussain
University of Illinois at Urbana-Champaign NUCES, Lahore
[email protected] [email protected]

Abstract- Urdu is a widely used language in South 2000, Encyclopedia of Writing and [2]). Urdu
Asia and is spoken in more than 20 countries. In has also retained its Persio-Arabic influence in
writing, Urdu is traditionally written in Nastaliq the form of the writing style or typeface. Urdu is
script. Though this script is defined by well-formed written in Nastaliq, a commonly used
rules, passed down mainly through generations of
calligraphers, than books etc, these rules have not
calligraphic style for Persio-Arabic scripts.
been quantitatively examined and published in Nastaliq is derived from two other styles of
enough detail. The extreme context sensitive nature Arabic script ‘Naskh’ and ‘Taleeq’. It was
of Nastaliq is generally accepted by its writers therefore named Naskh-Taleeq which gradually
without the need to actually explore this shortened to “Nastaliq”.
hypothesis. This paper aims to show both. It first
performs a quantitative analysis of Nastaliq and
then explains its contextual behavior. This
behavior is captured in the form of a context
sensitive grammar. This computational model
could serve as a first step towards electronic
Typography of Nastaliq.

I. INTRODUCTION
Urdu is spoken by more than 60 million
speakers in over 20 countries [1]. Urdu is derived
from Arabic script. Arabic has many writing
styles including Naskh, Sulus, Riqah and Fig. 1. Urdu Abjad
Deevani. Urdu however is written in Nastaliq
script which is a mixture of Naskh and an old
obsolete Taleeq styles. This is far more complex II. POSITIONAL AND CONTEXTUAL FORMS
than the others. Arabic is a cursive script in which successive
Firstly, letters are written using a flat nib letters join together. A letter can therefore have
(traditionally using bamboo pens) and both four forms depending on its location or position
trajectory of the pen and angle of the nib define a in a ligature. These are isolated, initial, medial
glyph representing a letter. Each letter has and final forms. Consider the following table 1,
precise writing rules, relative to the length of the in which letter ‘bay’ indicated in gray has a
flat nib. Secondly, this cursive font is highly different shape when it occurs in a) initial, b)
context sensitive. Shape of a letter depends on medial, c) final and d) isolated position. Since
multiple neighboring characters. In addition it Urdu is an derived from Arabic script and
has a complex mark placement and justification Nastaliq is used for writing Urdu, both Urdu and
mechanism. This paper examines the context Nastaliq inherit this property.
sensitive behavior of this script and presents a
context sensitive grammar explaining it. TABLE 1
POSITIONAL FORMS FOR LETTER BAY
A. Urdu Script
The Urdu abjad is a derivative of the Persian
alphabet derived from Arabic script, which in
itself is derived from the Aramaic script (Encarta
 Urdu, only half of them need to be looked into.
Letters ‘alif’, ‘dal’, ‘ray’ and ‘vao’ only have Note that only the characters that are used in
two forms. These letters cannot join from front place of multiple similar shapes are shown. The
with the next letter and therefore do not have an rest of the characters in the abjad are used
initial or medial forms. without any such similar-shape classification.
Nastaliq is far more complex than the 4-shape
phenomenon. In addition to position of character
TABLE 4
in a ligature, the character shape also depends on GROUPING OF LETTERS WITH SIMILAR BASE FORM
other characters of the ligature. Thus Nastaliq is Similar Base Forms Letter
inherently context sensitive. Table 2 below
shows a sample of this behavior in which a letter
bay, occurring in initial form in all cases, has
three different shape indicated in grey. This Also ‫ ن‬and ‫ ی‬in initial and
context sensitivity of Nastaliq can be captured by medial form
substitution grammar. This is discussed in detail
later in this paper.

TABLE 2
CONTEXTUAL FORMS FOR LETTER BAY IN INITIAL FORM

III. GROUPING OF ‘SIMILAR’ LETTERS

There are some letters in Arabic script and
consequently in Urdu, that share a common base
form. What they differ by is a diacritical mark
placed below or above the base form. This can
be seen in table 3 below which shows letters
‘bay’, ‘pay’, 'tay’, ‘Tey’ and ‘say’ in isolated
forms. And it’s clearly evident that they all have IV. METHODOLOGY: TABLETS
the same base form. This is also true for initial,
medial and final forms of these letters. The Nastaliq alphabets for Urdu have been
adapted from their Arabic counterparts as in the
TABLE 3 Naskh and T’aleeq styles from which it has been
LETTERS WITH SIMILAR BASE FORM derived. However, even for Urdu, this style is
still taught with its original alphabet set. When
the pupil gains mastery of the ligatures of this
alphabet, then he/she is introduced to the
modifications for Urdu.
The methodology employed for this study is
similar to how calligraphy is taught to freshmen.
The students begin by writing isolated forms of
letters. In doing so, they must develop the skill to
write a perfect shape over and over again by
Since these letters have a similar base shape, it maintaining the exact size, angle, position etc.
would be redundant to examine the shape of all When the students have achieved the proficiency
these letters in different positions and context. in isolated form it is said that they have
Studying the behavior of one letter would suffice completed the first ‘taxti’, meaning tablet. The
the others. Table 4 below shows all groupings first tablet is shown in figure 2 below; ‘taxti’ or
that are possible. The benefit of this grouping is tablet can be considered as a degree of
that instead of examining about 35 letters in excellence. First tablet is considered level 0
 Fig. 4. Jeem Tablet

Note: If the students learn to write ligatures

beginning with ‘bay’, then it is automatically
assumed they can also write ligatures beginning
with ‘tay’, ‘say’, etc because all these letters
have the same shape as ‘bay’ and only differ in
the diacritical mark. Please refer to section 3
above for details.
The level 0 and 1 helps to understand a lot
about the initial and final forms of letters. There
are however no further (formal) levels of any
kind. So, in addition to understanding the medial
Fig. 2. Tablet for Isolated forms forms, the students are also expected to learn to
write three and above letter ligatures on their
Once this level is completed, the students own through observation and consultation.
move over to level 2. In this level, they learn to For this study, a calligrapher was consulted
write all possible two-letter ligatures. This phase who provided the tablets and other text that was
is organized into 10 tablets. The first tablet required.
,,
consists of all ligatures beginning with ‘bay’, the
second abjad of Arabic script. For example in V. SEGMENTATION
English it would be like writing ba, bb, bc etc. all
the way till bz. Note that the first letter in Arabic Before moving over to the analysis of
script is ‘alif’ which has no initial form and Contextual substitution of Nastaliq, one of the
therefore does not form two letter ligatures predecessor worth mentioning is segmentation.
beginning with it. See section 2 above. The bay Since Nastaliq is a cursive script, segmentation
tablet is shown in figure 3 below. plays an important role in determining the
distinct shapes a letter can acquire. Consider a
two letter ligature composed of jeem and fay.
There can be a number of places from where this
ligature can be segmented (as indicated by 1,2
and 3 ) giving different shapes for initial form of
jeem.

(1) (2) (3)

If (1) is adopted as the segmentation scheme

and being consistent with it, when this is applied
Fig. 3. Bay Tablet on another two letter ligature that also begins
In the similar way the students move over to with a ‘jeem’, it follows that both the ligatures
the second tablet of this phase which has share a common shape of ‘jeem’. Both the
ligatures beginning with ‘jeem’ and so on. The ligature are shown below with segmentation
‘jeem’ tablet is also shown below in figure 4. approach (1). The resulting, similar ‘jeem’
shapes is also given.

Where as if the second (2) option is accepted for

segmentation, then the two initial forms will be
different.
 Initial-Position characters:
The following table lists the initial shapes of
letter ‘Bay’. The last eight shapes were
identified during analysis of three-character
ligature and do not realize in two character
The segmentation approach selected for this ligatures. Another way of putting it is that these
analysis is the later one or approach (2). The form can only occur in ligatures of length greater
reason for this selection is that the resulting than or equal to three. There is no such
shapes are a close approximation to the shapes in restriction for the first 16 shapes.
the calligrapher’s mind. Most probably because
these shapes represent complete stokes (, though TABLE 5
an expert calligrapher may be able to write the INITIAL FORMS OF BAY
whole ligature in one stroke). While in former
case there seems to be some discontinuity in the
smooth stroke of a calligrapher’s pen.
The discussion on contextual shape analysis in ‫ب‬Init1 ‫ب‬Init2 ‫ب‬Init3 ‫ب‬Init4
next section is therefore based on the “stroke
segmentation’ approach.

VI. CONTEXTUAL ANALYSIS OF NASTALIQ

‫ب‬Init5 ‫ب‬Init6 ‫ب‬Init7 ‫ب‬Init8
This section lists the context sensitive
grammar for characters occurring in initial
position of ligatures.

Explanation of Grammatical Conventions: ‫ب‬Init9 ‫ب‬Init10 ‫ب‬Init11 ‫ب‬Init12

The productions such as:

‫ ب‬Æ ‫ب‬1 / __<A> | ‫ ف‬Initial Form __

is to be read as ‫ ب‬transforms to ‫ب‬1 (‫ ب‬Æ ‫ب‬1), in ‫ب‬Init13 ‫ب‬Init14 ‫ب‬Init15 ‫ب‬Init16

the environment (/) when ‫ ب‬occurs before class
A (__<A>) or ( | ) when ‫ ب‬occurs before
initially occurring ‫ف‬.
Note that ‘OR’ ( | )operator has a higher ‫ب‬Init21 ‫ ب‬Init22 ‫ ب‬Init23 ‫ب‬Init24
precedence than ‘Forward Slash’ (/) operator.
Thus, it would be possible to write multiple
transformations in one environment using
several ‘OR’ ( | ) operator on the right side of a ‫ب‬Init25 ‫ب‬Init26 ‫ب‬Init27 ‫ب‬Init28
single (/).
The context in which these shapes occur has
Contextual Shift
been formulated in the form of a context
One invariant that have predominantly existed
sensitive grammar. The grammar for initial
in this contextual analysis of Nastaliq is that the
forms of ‘bay’ is given below. Note: The word
shape of a character is mostly dependent on
‘medial’ used in the grammar represents medial
immediate proceding character. That is given a
shapes of ‘bay’ given in appendix A
ligature composed of character sequence X1, X2
….XN for N>2, the shape of character Xi where i
< N, is determined by letter Xi+1. While all ‫ ب‬Æ ‫ ب‬Init1 / __ ‫__ ل‬ / ‫ﮎ‬ __ | ‫ا | __ د‬
preceding letters X1 ....Xi-1 and character ‫ ب‬Æ ‫ ب‬Init2 / __ ‫ب‬Final
sequences after its following character i.e Xi+2 ‫ ب‬Æ ‫ ب‬Init3 / __ ‫ج‬
…. XN have no (or little) role in its shaping. ‫ ب‬Æ ‫ ب‬Init4 / __ ‫ر‬
Sequence of bay’s form an exception to this ‫ ب‬Æ ‫ ب‬Init5 / __ ‫س‬
general rule. Other exceptions are also
mentioned
‫ ب‬Æ ‫ ب‬Init6 / __ ‫ص‬
‫ ب‬Æ ‫ ب‬Init7 / __ ‫ط‬
‫ ب‬Æ ‫ ب‬Init8 / __ ‫ع‬ Same
as
‫ ب‬Æ ‫ ب‬Init9 / __ ‫ف‬ shape
‫ ب‬Æ ‫ ب‬Init10 / __ ‫و ___ | ق‬ 11
‫ ب‬Æ ‫ ب‬Init11 / __ ‫م‬ ‫ج‬Init25 ‫ ج‬Init26 ‫ ج‬Init27 ‫ج‬Init28
‫ ب‬Æ ‫ ب‬Init12 / __ ‫ن‬
‫ ب‬Æ ‫ ب‬Init13 / __ ‫ﮦ‬Final Note that the terminating shape of ‫ ب‬-init2 is
very different from that of ‫ج‬-init2. This is
‫ ب‬Æ ‫ ب‬Init14 / __ ‫ه‬
because they connect to different shapes of final
‫ ب‬Æ ‫ ب‬Init15 / __ ‫ﯼ‬ ‘bay’. This is discussed in the next section.
‫ ب‬Æ ‫ ب‬Init16 / __ ‫ے‬
‫ ب‬Æ ‫ ب‬Init22 /__‫ ب‬Medial Only if none of the below Final Forms:
apply With the exception of ‘bay’ and ‘ray’, all other
‫ ب‬Æ ‫ ب‬Init21 /__‫ ب‬Medial2 letters have a unique final form. Both ‘bay’ and
‫ ب‬Æ ‫ ب‬Init23 /__‫ ب‬Medial4 ‘ray’ have two final forms. These are shown in
the table below. For ‘bay’, the first form occurs
‫ ب‬Æ ‫ ب‬Init24 /__‫ ب‬Medial5
when it is preceded only by ‘bay’, (shape ‫ ب‬-
‫ ب‬Æ ‫ ب‬Init25 /__‫ ب‬Medial13 init2 in the table above), ‘fa’, ‘qaf’, ‘la’ and ‘ka’,
‫ ب‬Æ ‫ ب‬Init26 /‫__ م ا | __ م د‬ all in initial forms. While the other more frequent
/ __‫ __ م ل‬/ ‫م ﮎ‬ is realized else where and an example of letter
connecting to it is ‘jeem’ having the form ‫ج‬-
‫ ب‬Æ ‫ ب‬Init27 /__‫ م‬Otherwise init2. This explains the noticeable difference
‫ ب‬Æ ‫ ب‬Init28 /__‫ ﮦ‬Medial between the ending strokes of ‫ ب‬-init2 and ‫ج‬-
init2.
All other letters have similar number of shapes TABLE 7
occurring in similar context. The initial forms of FINAL FORMS FOR LETTER BAY AND RAY
‘jeem’ have been shown in table below. The
grammar of ‘jeem’ can be derived from grammar
of ‘bay’. Likewise, shapes of other letters can be ‫ب‬Final 1 ‫ ب‬Final 2
deduced from table 5 and table 6. Medial Shapes
of Bay have been listed in Appendix A
TABLE 6 ‫ ر‬Final 1 ‫ ر‬Final 2
INITIAL FORMS OF JEEM

The context for each of these shapes is given

below. Interesting observation here is that ‘ray-
final1’ occurs only when it is preceded by initial
‫ج‬Init1 ‫ج‬Init2 ‫ج‬Init3 ‫ج‬Init4 forms of ‘bay’ and ‘jeem’. There is however no
such precinct for letters ‘ka’ and ‘la’, which can
be in any form (initial or medial).

‫ج‬Init5 ‫ج‬Init6 ‫ج‬Init7 ‫ج‬Init8 ‫ ب‬Æ ‫ ب‬Final 1 / ‫ ب‬Initial Form __

‫ ب‬Æ ‫ ب‬Final 1 / ‫ ف‬Initial Form __
‫ ب‬Æ ‫ ب‬Final 1 / ‫ ق‬Initial Form __
‫ ب‬Æ ‫ ب‬Final 1 / ‫ ل‬Initial Form __
‫ج‬Init9 ‫ج‬Init10 ‫ج‬Init11 ‫ج‬Init12 ‫ ب‬Æ ‫ ب‬Final 1 / ‫ ﮎ‬Initial Form __
‫ ب‬Æ ‫ ب‬Final 2 otherwise
‫ ر‬Æ ‫ ر‬Final 1 / ‫ ب‬Initial Form __
‫ ر‬Æ ‫ ر‬Final 1 / ‫ ج‬Initial Form __
‫ج‬Init13 ‫ج‬Init14 ‫ج‬Init15 ‫ج‬Init16 ‫ ر‬Æ ‫ ر‬Final 1 / ‫__ ﮎ‬
‫ ر‬Æ ‫ ر‬Final 1 / ‫__ ل‬
‫ ر‬Æ ‫ ر‬Final 2 otherwise
‫ج‬Init21 ‫ ج‬Init22 ‫ ج‬Init23 ‫ج‬Init24
 ACKNOWLEDGEMENT
The work presented has been partially supported by REFERENCE
a Small Grants Program by IDRC, APDIP UNDP
and APNIC. The work on Nastaliq has been [1] www.ethnologue.com
completed by the Nafees Nastaliq team at CRULP, [2] http://en.wikipedia.org/wiki/Aramaic_script
NUCES.

APPENDIX A: MEDIAL SHAPES OF BAY WITH EXAMPLES

Given below is a detailed analysis of 20 medial shapes and 4 initial shapes of letter ‘bay’. These 4 initial
shapes are mentioned here since they are used in medial position of a ligature also. Following table lists
context sensitive grammar of a particular shape and the shape’s glyph it self when it occurs in medial
position. Also included is an example corresponding to each glyph.
‫ ب‬Æ ‫ ب‬Medial1 / __<A> ‫ ب‬Æ ‫ ب‬Medi13 / __ ‫م‬
| __ ‫ل‬
| __ ‫ب‬Init2 ‫ ب‬Medi13
| __ ‫ب‬Init6 ‫ب‬Medial1
| __ ‫ ب‬Init14
| __ ‫ ب‬Medial23 ‫ ب‬Æ ‫ ب‬Init14 / __ ‫ن‬Final
| __ ‫ ب‬Medial25
‫ ب‬Æ‫ ب‬Init2/ __‫ ب‬Medi1 ‫ ب‬Init14
| __ ‫ ب‬Medi2 ‫ ب‬Æ ‫ ب‬Init15 / __ ‫ﮦ‬
| __ ‫ ب‬Medi8
| __ ‫ ب‬Medi9
‫ ب‬Init2
| __ ‫ ب‬Medi10
| __ ‫ ب‬Medi12
‫ ب‬Init15
‫ ب‬Æ ‫ ب‬Medi16 / __ ‫ه‬
| __ ‫ ب‬Init15
| __ ‫ ب‬Medi16
| __ ‫ ب‬Medi18
‫ ب‬Medi16
‫ ب‬Æ ‫ ب‬Medi2 / __ ‫ب‬Final ‫ ب‬Æ ‫ ب‬Medi17 / __ ‫ﯼ‬

‫ب‬Medi2 ‫ ب‬Medi17
‫ ب‬Æ ‫ ب‬Medi3 / __ ‫ج‬
‫ ب‬Æ ‫ ب‬Medi18 / __ ‫ے‬
‫ب‬Medi3
‫ ب‬Medi18
‫ ب‬Æ ‫ ب‬Medi5 / __ ‫ر‬Final2 ‫ ب‬Æ ‫ ب‬Medi23/ __‫ ب‬Medi3

‫ ب‬Medi23
‫ب‬Medi5
‫ ب‬Æ ‫ ب‬Init6 / __ ‫س‬
‫ ب‬Æ ‫ ب‬Medi25/ __‫ ب‬Medi5

‫ ب‬Init6 ‫ ب‬Medi25
‫ ب‬Æ ‫ ب‬Medi8 / __ ‫ص‬

‫ ب‬Medi8 ‫ ب‬Æ ‫ ب‬Medi33/ __‫ ب‬Medi13

‫ ب‬Æ ‫ ب‬Medi9 / __ ‫ط‬ ‫ ب‬Medi33
‫ ب‬Medi9
‫ ب‬Æ ‫ ب‬Medi10 / __ ‫ع‬

‫ ب‬Medi10 ‫ ب‬Æ ‫ ب‬Medi36/ __‫ ب‬Medi16

‫ب‬Æ ‫ ب‬Medi11 / __ ‫ف‬
‫ ب‬Medi36
‫ ب‬Medi11
‫ ب‬Æ ‫ ب‬Medi12 / __‫ق‬
‫ ب‬Æ ‫ ب‬Medi37/ __‫ ب‬Medi17
___ ‫و‬
‫ ب‬Medi12 ‫ ب‬Medi37