A Diachronic Approach for Schwa Deletion in Indo Aryan Languages 
Monojit CHOUDHURY, Anupam BASU and Sudeshna SARKAR 
Dept.. of Computer Science & Engineering, 
Indian Institute of Technology, Kharagpur 
INDIA, PIN-721302 
{ monojit, anupam, sudeshna } @cse.iitkgp.ernet.in  
 
 
Abstract 
Schwa deletion is an important issue in 
grapheme-to-phoneme conversion for Indo-
Aryan languages (IAL). In this paper, we 
describe a syllable minimization based 
algorithm for dealing with this that 
outperforms the existing methods in terms of 
efficiency and accuracy. The algorithm is 
motivated by the fact that deletion of schwa is 
a diachronic and sociolinguistic phenomenon 
that facilitates faster communication through 
syllable economy. The contribution of the 
paper is not just a better algorithm for schwa 
deletion; rather we describe here a constrained 
optimization based framework that can partly 
model the evolution of languages, and hence, 
can be used for solving many problems in 
computational linguistics that call for 
diachronic explanations. 
 
1 Introduction 
Linguists propose new models for languages 
in order to explain language acquisition and 
processing by humans. Irregularities and 
exceptions to the theories are often explained by 
evidence from diachronic linguistics and other 
social and external phenomena. Absence of 
diachronic analysis in computational modelling of 
languages results in a large number of exceptions, 
which are commonly handled by ad hoc rules or 
exhaustive enumeration. These techniques lead to 
poor scalability and lack of graceful degradation of 
the systems along with increased complexity. 
Although complete modelling of the evolution of 
language is impossible due to the involvement of 
myriads of socio-political and cultural factors, it is 
definitely possible to model certain basic 
principles of language change. 
In this paper we describe an algorithm for 
schwa deletion in Indo-Aryan Languages (IAL) 
that is motivated by the diachronic evolution of the 
languages. The proposed computational framework 
models languages as a constrained optimization 
system, where a language evolves by optimizing 
the rate of communication, subjected to a set of 
constraints such as ease of articulation and 
learning, and acoustic distinctiveness. A syllable 
minimization based optimization function fitted to 
the aforementioned model has been used for 
solving the problem of schwa deletion with 
considerable success.  
The paper is organized as follows: Section 2 
defines the problem and discusses some of the 
previous works. Section 3 describes the current 
models of language evolution, which has been 
used to develop a computational framework 
described in the next section. Section 5 and 6 
presents the algorithm and its experimental 
analysis respectively. Section 7 concludes the 
paper summarizing our contributions. 
2 The Problem  
Schwa is defined as the mid-central vowel that 
occurs in unstressed syllables. The first vowel of 
the IAL alphabet {a}
1
 is the schwa. Normally, it is 
pronounced as /ə / in Hindi and Sanskrit, and as /ɔ / 
in Bengali. Schwa deletion is a phonological 
phenomenon where schwa is absent in the 
pronunciation of a particular word, although 
ideally it should have been pronounced (Ohala, 
1983).  
Sanskrit and some of the modern IAL that have 
evolved from it (e.g. Hindi and Bengali), are 
written from left to right using Brahmi-derived 
scripts. All the vowels are explicitly represented 
using diacritical or non-diacritical marks around 
the consonant except for the schwa, which is the 
inherent vowel. Unlike Sanskrit, many modern 
IAL like Hindi and Bengali allow deletion of 
schwa in certain contexts. Table I illustrates this 
phenomenon for the three languages. In order to 
determine the proper pronunciation of the words, it 
is necessary to predict which schwas are deleted 
and which are not. Thus, schwa deletion is an 
                                                      
1
 The graphemes for Indo-Aryan languages are written within ‘{’ and ‘}’ 
according to the scheme adopted by the International Congress of Orientalists at 
Athens in 1992. The phonetic transcriptions are written within two ‘/’ using the 
IPA symbols. 
                                                                  Barcelona, July 2004
                                              Association for Computations Linguistics
                       ACL Special Interest Group on Computational Phonology (SIGPHON)
                                                    Proceedings of the Workshop of the
important issue for grapheme-to-phoneme 
conversion of IAL, which in turn is required for a 
good Text-to-Speech synthesizer (Narasimhan et 
al, 2001).  
 
Pronunciation   The  
Spelling Sanskri
t 
Hind
i 
Bengali 
s¡phaly
a 
(succes
s) 
saφə ly
ə  
(3) 
saφə
ljə  
(3) 
ʃ aφol lo 
(3) 
racan¡ 
(creati
on) 
rə cə na 
(3) 
rə cn
a 
(2) 
rɔ cona 
(3) 
veda 
(Veda) 
vedə  
(2) 
ved 
(1) 
bed 
(1) 
 
Table 1. Pronunciation of three different words 
in three different IAL. The number of syllables is 
denoted within parenthesis below the 
pronunciations. In Bengali {a} can also be 
pronounced as /o/ in certain contexts. 
 
Several theories have been proposed on the 
linguistic aspects of schwa deletion in Hindi (Pray, 
1970; Kaira 1976; Ohala, 1977, 1983) and its 
diachronic evolution (Misra, 1967). Ohala (1983) 
has summarized the rule for schwa deletion in 
Hindi as  
ə  →   φ / VC __ CV 
Condition 1: There may be no morpheme 
boundary in the environment to the left. 
Condition 2: The output of the rule should not 
violate the phonotactic constraints of Hindi 
Convention: The rule applies from right to left 
 
The explanation of the rule was based on 
psycholinguistic evidence; diachronic facts were 
used only to explain the exceptions. Narsimhan et 
al (2001) designed an algorithm for schwa deletion 
in Hindi based on this work. The reported accuracy 
of the algorithm is 89%. Some rules for word final 
schwa deletion in Bengali have been proposed by 
Chatterji (1926), but we do not know of any work 
on computational modelling. 
 
3 Factors governing language change 
The fact that schwa deletion in IAL is a diachronic 
phenomenon has been substantiated by Misra 
(1967). According to Ohala (1983) the deletion of 
schwas is more frequent in casual and fast speech 
compared to formal and slower ones. It can be 
inferred from these facts that the motivation behind 
schwa deletion is faster communication through 
minimization of syllables (Tranel 1999).  
 Some recent works on mathematical and 
simulation based modelling of language evolution 
(Boer, 2000; Cangelosi and Parisi, 2002; Nowak et 
al, 2002) suggests that several features of 
languages emerge due to some basic cognitive and 
articulatory factors. These models assume a) ease 
of articulation, b) ease of learning, and c) acoustic 
distinctiveness as the primary driving forces 
behind language evolution. The three forces 
operate simultaneously over the language in order 
to maximize the rate of successful communication 
in terms of time and effort spent by the language 
users to generate, understand and learn the 
language. Thus, language can be modelled as a 
multi-objective optimization system, where the 
optimization criteria are 
 
• Minimization of effort (in terms of energy and 
time spent while conveying a piece of 
information)  
• Minimization of learning time and effort 
• Minimization of probability of 
misunderstanding (in the sense of confusing 
one word with another) 
 
These three criteria are mutually contradictory and 
therefore there exists no global optimum. Let us 
examine the phenomenon of schwa deletion under 
this multi-objective optimization model for 
language evolution. When a vowel is deleted from 
a word the number of syllables reduces by one. For 
example, in Table 1, for the second word, Sanskrit 
and Bengali have three syllables, whereas due to 
the deletion of a schwa, the Hindi pronunciation 
has only two syllables. Reduction of syllables 
implies shorter time for pronunciation of a word, 
and hence faster communication. However, 
deletion of schwas in certain contexts might result 
in a consonant cluster which the native speakers 
find very difficult or impossible to pronounce. This 
beats the very purpose of schwa deletion, i.e. the 
minimization of effort of articulation and therefore, 
is unacceptable. The second condition for the rule 
proposed by Ohala (section 2) refers to this 
constraint. 
 There are contexts where deletion of schwa 
would not give rise to inadmissible consonant 
clusters. For example, in the Hindi/Bengali word 
pari (fairy, /pə ri/ in Hindi), if the first schwa is 
deleted, the pronunciation would be /pri/, which 
does not violate the phonotactic constraints of the 
languages. The schwa, however, is not deleted, 
because /pə ri/ and /pri/ are too distinct from each 
other to be interpreted as the same word. 
Moreover, /pri/ is closer to other Hindi words like 
priya (favorite, /prijə /). In this case, the deletion of 
schwa reduces the acoustic distinctiveness of the 
word from other words in the lexicon, which 
increases the probability of misunderstanding, and 
hence the schwa might not be deleted in such a 
context. 
4 Computational framework 
We propose the following diachronic 
explanation for schwa deletion in IAL.  
In old IAL none of the schwas are deleted. The 
modern IAL use the script and spelling 
conventions similar to Sanskrit. Due to a higher 
evolutionary pressure on the spoken forms of the 
languages than on the written forms, schwas are 
deleted in the pronunciation, but are still present in 
the graphemic forms. The deletion is a slow 
diachronic phenomenon, where in order to 
communicate faster, initially the speakers 
unintentionally deleted the schwas. Only those 
deletions were acceptable that did not lead to a 
syllable structure which was too difficult to 
pronounce, learn or understand for the native 
speakers. Gradually, the pattern of deletion spread 
across the population and over the different items 
in the lexicon. 
In this section, we describe a computational 
framework for modelling the aforementioned 
hypothesis based on the three optimization criteria 
stated in the last section. The aim of the proposed 
framework is not to validate the hypothesis 
through micro-simulation (Cangelosi and Parisi, 
2002); rather it tries to predict the schwa deletion 
pattern based on the optimizations that might have 
affected the deletion of schwas diachronically. In 
the next section, we present an efficient algorithm 
for schwa deletion in IAL, which can be 
automatically constructed from this model, without 
the help of any other evidence. 
 
4.1 Basic definitions 
 All the unexplained symbols used below stand 
for their usual meaning in the context of formal 
language theory. Please refer to (Hopcroft and 
Ullman, 1979) for details. 
 
Σ
g
 (Σ
p
): A finite set of the graphemes
2 
(phonemes) in the language  
Σ
g
 = V
g
 ∪ C
g
,  Σ
p
 = V
p
  ∪ C
p
  
                                                      
2
 Graphemes here do not refer to glyphs. Free vowels and their 
corresponding diacritical marks are considered to be the same symbol 
Where 
V
g
 (V
p
): Finite set of graphemes (phonemes), 
which are vowels 
C
g
 (C
p
): Finite set of graphemes (phonemes), 
which are consonants. Semivowels are also 
considered as consonants. 
 
α  ∈ V
g
 is a special symbol that represents schwa. 
We define, 
f
g2p
: Σ
g
 → Σ
p
 
f
g2p
 is the default mapping of the graphemes to 
the phonemes. This oversimplification is made 
here for two reasons. First, since IAL use a 
phonetic script, this in general is true
3 
and second, 
this assumption does not have any affect on the 
schwa deletion algorithm.  
 
A word w is defined as a 2-tuple <w
g
, w
p
>, 
where  
w
g 
∈ Σ
g
+
 and w
p
∈ Σ
p
+ 
 
A lexicon Λ  is the union of all the valid words w 
of a language. A grapheme-to-phoneme converter 
is defined as a function F
g2p
: Σ
g
+
→ Σ
p
+
, such that 
∀w < w
g
, w
p
 >∈ Λ , F
g2p
(w
g
) = w
p 
 
4.2 Phonotactic constraints 
In order to model the ease of articulation, we 
start with the modelling of phonotactic constraints. 
A consonant cluster is a string of the form C
p
C
p
+
. 
Phonotactic constraints restrict the presence of 
some of the consonant clusters in the phonetic 
representation of a word (w
p
). At the most generic 
level we can think of a consonant cluster ranking 
(CCR) function, where ℕ is the set of natural 
numbers       
                 
CCR: C
p
+
 → ℕ  
 
The function CCR is independent of any 
language and every language has a threshold τ
CCR
, 
such that a consonant cluster x ∈ C
p
+
 is allowed in 
the language if and only if    
           
CCR (x) ≤ τ
CCR 
 
We define two special variants of CCR, O_CCR 
and C_CCR
, 
which ranks the admissibility of the 
consonant clusters at the onset and coda positions 
respectively. The definition is similar to that of 
CCR, and τ
OCCR
 and τ
CCCR
 are the corresponding 
threshold values.  
                                                      
3
 This assumption is not strictly valid since a cluster of consonant might be 
mapped to a single consonant or a different cluster.    
The sonority hierarchy (Vennemann, 1988) and 
markedness conditions (Kager, 1999) along with 
the physiology of the articulatory mechanism point 
towards the existence of a language independent 
ranking function as hypothesized above. However, 
there might be accidental gaps in the list of 
admissible consonant clusters of a language 
(Ohala, 1983), which can not be explained on the 
basis of CCR alone. Therefore, we define a 
Boolean function ADM that tell us about the 
admissibility of consonant clusters in a language. 
 
ADM: C
p
+
 → {0, 1}, such that for s ∈ C
p
+
 
 (ADM (s) = 1) ⇔  (s is an admissible cluster) 
 
In general, we can derive this function from 
CCR as 
ADM (s) = sign (τ
CCR    
– CCR (s)) 
 
However, we might have to forcefully convert 
some values to 0 due to accidental gaps.   
 
4.3 Syllable and Syllabification 
We define a syllable σ as a regular expression, 
with the assumption that the nucleus contains a 
single vowel. Thus, 
σ ∈ C
p
* V
p
 C
p
* 
 
The syllabification function SYL maps the 
phonetic representation w
p
 of a word w to a string 
of syllables σ
1
σ
2
…σ
m
 such that the effort of 
articulation and learning are minimum. 
 
We model the effort of articulation using a 
syllable ranking function SR, which is similar to 
CCR. 
 SR
 
: C
p
*V
p
C
p
* → ℕ  
 
SR is mainly dependent on the structure of the 
syllable. We enumerate the first few terms of the 
function SR.          
 
SR (V
p
) = 1,  SR (C
p
V
p
) = 2   
SR (C
p
V
p
C
p
) = 3, SR (V
p
C
p
) = 4 
SR (C
p
C
p
V
p
) = 5,  SR (C
p
C
p
V
p
C
p
) = 6 
SR (C
p
C
p
C
p
V
p
) = 7,  SR (C
p
V
p
C
p
C
p
) = 8 
 
For all other possible syllable structures σ’,  
SR (σ’) > 8 
 
Also, for any syllable σ
, 
   
[O_CCR (onset(σ)) >  τ
OCCR
] ∨ 
[C_CCR (coda(σ)) >  τ
CCCR
] ⇒ (SR (σ) = ∞ ) 
 
This means that if either the coda or the onset of 
a syllable is inadmissible, then the ranking 
function maps the syllable to the highest possible 
rank, represented symbolically by the infinity (∞). 
onset and coda are projection functions that project 
the longest valid prefix and suffix of a syllable 
respectively that are elements of  C
p
*. 
 
We define a syllabification to be valid if all the 
syllables are valid (i.e. strings of the form 
C
p
*V
p
C
p
*) and every symbol in the word is a part 
of one and only one syllable in the syllabification. 
We can define a partial ordering, ≤
σ, 
among the 
possible valid syllabifications of a given word 
based on SR
p
 such that the syllabification with 
smaller number of high ranked syllables is 
preferred to one that has more hard (high ranked) 
syllables. Now we define SYL (w
p
) as the set of all 
possible syllabifications σ
1
σ
2
…σ
m 
such that (i) 
σ
1
σ
2
…σ
m 
is a valid syllabification of
 
 w
p
 and (ii) 
there exist no other valid syllabification v of w
p 
such that v ≤
σ 
σ
1
σ
2
…σ
m
.  
 
The definitions of syllable and syllabification are 
motivated by the markedness conditions (Kager, 
1999) and experimental results on child language 
acquisition (MacNeilage and Davis, 2000), that 
show that some syllables and syllabifications are 
easier to learn and pronounce than others. 
 
4.4 Acoustic distinctiveness constraints 
Perceptual experiments show that speakers 
always articulate the onset of the syllables more 
clearly and correctly compared to the articulations 
of the vowel and the coda (Fosler-Lussier et al, 
1999; Greenberg, 1999). Therefore, it is likely that 
the hearer distinguish between syllables by paying 
more weight to the onset than to the coda. A 
continuous distance metric D
σ 
might be defined 
based on these experimental results, such that the 
probability of confusion (interpreting one syllable 
as another) between two syllables σ and σ’ 
increases as the value of D
σ
(σ , σ’) decreases. We 
can further define an acoustic distance function D
w
 
using the function D
σ 
, which measures the 
probability of confusion between two arbitrary 
words in the phonetic domain. 
In the case of schwa deletion, however, we want 
the acoustic distance between the ideal 
pronunciation (without any schwa deletion) and 
the normal pronunciation (with schwa deletion) to 
be smaller, so that the word is not confused with 
other words in the lexicon. Formally, for the 
graphemic representation of a word w
g
 = x
1
x
2
 … 
x
n
, 
D
w
(f
g2p
 (x
1
).f
g2p
 (x
2
)… f
g2p
 (x
n
), F
g2p
(w
g
)) < 
τ
critical
, where τ
critical
 is the maximum allowable 
distance and ‘.’ is the concatenation operator. 
Rather than modelling this as an optimization 
criterion, we reformulate this as a constraint. The 
simplification in this case serves our purpose. 
 
We define, where x ∈ C
p
   
D
σ
(x. f
g2p
(α ), φ) = 0                  (4a) 
 D
σ
(σ, σ.x) = 0          (4b) 
For all other cases D
σ  
is infinity (∞), unless the 
two syllables are identical.             (4c) 
 
(4a) allows the deletion of a schwa from an open 
syllable; (4b) allows the concatenation of a 
consonant at the coda position. This is motivated 
by the fact that coda has least distinctiveness 
(Greenberg, 1999). (4c) restricts any change at the 
onset of a syllable or the vowels other than schwa.  
 
On the basis of D
σ
 we can define 
 
D
w
(w
p1
, w
p2
) = 
0 if and only if there exists an alignment between 
the sequences SYL (w
p1
) and SYL (w
p2
), with 
possible gaps (φ or null syllables) such that for all 
the corresponding pairs of syllable taken from the 
two sequences, the acoustic distinctiveness (D
σ
) 
between them is 0. Thus, only operations allowed 
are deletion of a schwa and addition of a consonant 
at the coda position. Anything else is forbidden for 
the sake of acoustic distinctiveness.  
 
We conclude this section by summarizing below 
the salient features of the model by comparing it 
with the optimization criteria stated in section 3. 
 
• The functions SR, CCR and its variants 
that rank the phonotactic constraints is a measure 
of the effort of articulation, learning and the 
probability of misunderstanding. Therefore we 
want to minimize it. However, it has been 
modelled as a constraint (ADM). 
• The function SYL is so defined that the 
efforts of articulation and learning are minimized. 
• D
w 
models the acoustic distinctiveness i.e. 
the criterion 3c, but it has been reformulated as a 
constraint as well.  
5 The algorithm 
We want to define F
g2p 
for a language given 
ADM and D
w
. F
g2p
 should be such that it enables 
faster communication by minimization of syllables 
by deletion of schwa. 
 
5.1 Formal definition 
Let w
g
 be an input sequence of graphemes to the 
function F
g2p
. Let w
p
 ∈ Σ
p
* be obtained by 
replacing all graphemes x in w
g
 by f
g2p
(x). Let w
p
’ 
be obtained by deletion of some (possibly all or 
none) of the schwas (f
g2p
(α)) in w
p
. F
g2p
(w
g
) = w
p
’, 
if and only if D
w
(w
p
, w
p
’) = 0  and (∀ v
p
)[( v
p
 can 
be obtained by deleting schwas from  w
p
) ∧ 
(D
w
(w
p
, v
p
) = 0) ⇒ |SYL
g
(w
p
’)| ≤ |SYL
g
(v
p
)| ] 
 
In words it means that among all w
p
’ obtainable 
by deletion of some of the schwas from w
p
, that 
respects both the ADM (phonotactic) and D
w
 
(acoustic distinctiveness) constraints, the one with 
the minimum number of syllables is chosen as the 
output of F
g2p
.  
 
procedure SYL : 
input: w
p
, O_CCR, C_CCR 
output:  σ
1
σ
2
…σ
m
  //The syllabification 
 
1. Include up to the first vowel in w
p
 in σ
1 
2. If there are 2 consonants c
1
c
2
 between the 
current vowel and the next vowel, include c
1
 in the 
current syllable and c2 in the next syllable. 
3. If there are 3 consonants c
1
c
2
c
3 
between the 
current vowel and the next vowel,  
     3.1 if O_CCR
p
(c
2
c
3
)≤ τ
OCCR 
, include c
1 
in the 
current syllable and c
2
c
3
 in the next syllable  
     3.2 else if C_CCR
p
(c
1
c
2
)≤ τ
CCCR
 include c
1
c
2 
in the 
current syllable and c
3
 in the next syllable 
     3.3 else NO syllabification is possible 
4. If there is one or no consonant between the 
current vowel and the next vowel, terminate the 
current syllable and begin the next syllable 
5. Continue from step 2 till there are symbols not 
included in any syllable.  
end procedure 
 
 
Figure 1. Algorithm for syllabification 
 
5.2 A greedy strategy  
Figure 1 describes a linear time algorithm for 
syllabification (SYL) that conforms to the 
definition provided in section 4.3. This uses the 
fact that the maximum length of allowable 
consonant clusters for IAL is three. After 
syllabification of w
p
, we try to greedily delete the 
schwas so that the constraints specified by 4a, 4b 
and 4c are not violated. 4a states that only a schwa 
which is a part of an open syllable (cα, where c ∈ 
C
p
) can be deleted and 4b states that after schwa 
deletion, 
the consonant c is appended to the coda of the 
previous syllable. Therefore, both of them together 
imply schwas in two consecutive syllables cannot 
be deleted. Along with that, the following 
constraints can also be derived from the D
w
 
constraints (the reasons are omitted due to space 
constraints): 
R1. Schwa of the first syllable cannot be deleted 
R2. Schwa cannot be deleted before a consonant 
cluster. 
R3. The word final schwa can always be deleted 
unless the appending of the penultimate 
consonant to the previous syllable results in an 
inadmissible cluster. 
R4. For Bengali, which does not allow complex 
codas, schwas cannot be deleted after 
consonant clusters. 
R5. A schwa followed by a vowel cannot be 
deleted. 
 
procedure F
g2p
: 
input: w
g
 , ADM 
output:  w
p
  //The pronunciation 
 
1. w
p
’  = f
g2p
(x
1
).f
g2p
 (x
2
)… f
g2p
 (x
n
), where w
g
 is 
<x
1
x
2
 … x
n 
> 
2. Syllabify w
p
’  using procedure SYL
 
3. Using rules R1 to R6 and ADM constraints mark 
the schwas which cannot be deleted as F 
4.  While traversing the word from right to left 
         4.1 Delete a schwa if it is not marked F 
         4.2 Appended the dangling consonant to the 
coda of the adjacent syllable (to the left) 
         4.3 If the adjacent syllable (to the left) has a 
schwa which is unmarked, mark it F 
         4.4 Go to 4.1 if there are more schwas to the left 
of the current position. 
5. At the end of step 4 we get the syllabified string of 
phonemes <x’
1
x’
2
 … x’
m 
>, which is the required 
output 
end procedure 
 
Figure 2. Algorithm for schwa deletion 
 
 We have the following rule that cannot be 
captured by the constraints: 
R6. Schwa following a y (pronounced as /j/) 
cannot be deleted if it is preceded by a high 
vowel because /j/ is a glide from high vowel to 
a low/medium vowel (schwa), deletion of 
schwa would make the presence of the glide 
imperceptible. 
This rule could have been captured by the D
σ
 
constraints but we state it here as a separate rule 
for the sake of simplicity. Figure 2 describes an 
algorithm for schwa deletion using the rules above. 
It is easy to see that the time complexity of the 
algorithm is O(|w
g
|). Due to limited space, we omit 
the proof that the algorithm for F
g2p
 indeed 
minimizes the number of syllables without 
violating the constraints specified by ADM and D
w
. 
However, there might be more than one (precisely 
2) possible solutions and in that case the algorithm 
chooses one of the solutions on the basis of the 
direction of traversal at step 4. The right to left 
traversal gives better results (as has been 
confirmed by Ohala, 1983) because the duration of 
syllables reduces towards the end of the word and 
hence the tendency to delete schwas at the word 
final position increases. 
6 Experimental Results and Discussions 
The algorithm was implemented for Bengali and 
Hindi and tested on a set of words. Table 2 
summarizes the results for Hindi (tested on the 
words in a pocket dictionary (Hindi-Bangla-
English, 2001)). The algorithm for Bengali was 
tested on 1000 randomly selected words from a 
corpus and found to be around 85% accurate.  
Some of the important features of the algorithm 
are as follows. 
• Efficiency: The algorithm runs in linear time 
on the input word length. It scans the whole 
word just twice. Thus, the hidden constant is 
also very small. 
• Polymorphemic Words: The algorithm can 
handle polymorphemic words, if the 
morphological information about the word is 
provided. This is because schwa deletion is not 
carried across morpheme boundaries. 
Morphological analyzer for Hindi and Bengali 
were implemented and integrated with the 
algorithm. For Hindi, the results were nearly 
perfect (99.89%) 
Exceptions: For Hindi there was hardly any 
exception to the algorithm. For Bengali, the types 
of words that were incorrectly processed by the 
algorithm include a class of very frequently used, 
disyllabic modifier adjectives, certain suffixes, 
borrowed words from Sanskrit and compound 
words. In Bengali, the schwa which is retained (as 
opposed to the predictions by the algorithm) are 
pronounced as /o/ and not as / ɔ /. Since, /o/ is not a 
central vowel, deletion of /o/ is marked as 
compared to deletion of / ɔ / which is unmarked. 
Transformation of schwa to some non-neutral 
vowel in Hindi is unknown and therefore, the 
algorithm works perfectly for Hindi. 
 
Experiment
al results for 
Hindi 
Test 
Size 
(No. of 
words) 
Incorre
ct results 
Accurac
y 
Without 
MA 
1109
5 
431 96.12% 
With MA 1109
5 
12 99.89% 
 
Table 2. Experimental results for Hindi schwa 
deletion. The results are for individual words. MA 
stands for Morphological Analysis 
7 Conclusion 
In this paper, we have described the 
phenomenon of schwa deletion in the IAL and 
proposed a diachronic explanation for it. In order 
to model the diachronic evolution, we used the 
concepts of ease of articulation, ease of learning 
and acoustic distinctiveness. We developed a 
computational framework, where we reformulated 
some of the optimization criteria as constraints and 
one of them (the syllable minimization) as the 
basic optimization function. The outcome of this is 
an efficient and accurate algorithm for solving 
schwa deletion in IAL. 
The contribution of this paper is not just a 
better algorithm for schwa deletion, which is 
necessary for developing Text-to-speech 
synthesizers for IAL, but a new approach based on 
a constrained optimization framework, motivated 
by the diachronic evolution of languages. A closer 
look at the algorithm will reveal that it is not much 
different from the schwa deletion rule proposed by 
Ohala (1983). However, Ohala’s rule was based on 
psycholinguistic and empirical observations, 
whereas we have derived the rule from a set of 
very basic assumptions (minimization of syllables 
and certain constraints). The algorithm itself can 
provide an explanation for the phenomenon. 
It must be mentioned that neither the aim nor 
the findings of this work are meant to propose a 
new model of language change. The models and 
concepts used here were all present previously and 
we have assumed and included some of them 
directly in our model. Our finding is not a proof of 
those models and can be considered only as a 
further validation. Our only claim here is that 
diachronic clues can help solve important 
problems in computational linguistics and for this 
we provide a computational framework and a 
specific example. 
 Some of the questions that we would like to 
address in the future include modelling of optional 
schwa deletion in Bengali compound words, 
evolution of morpho-phonology for Bengali verb 
systems, and modelling of dialect diversity using 
diachronic clues. More realistic, yet manageable 
computational frameworks for holistic or detailed 
modelling of language evolution can also be an 
interesting area of future research. 

References  
Bart de Boer 2000. Self Organization in Vowel 
Systems. Journal of Phonetics, 28:441-465 
Angelo Cangelosi and Domenico Parisi (Eds) 
2002. Simulating the Evolution of Language. 
Springer-Verlag, London  
Suniti K. Chatterji 1926. The Origin and 
Development of the Bengali Language. Rupa and 
Co. 
Eric Fosler-Lussier, Steven Greenberg and N 
Morgan  1999. Incorporating contextual 
phonetics into automatic speech recognition. 
Proc. Int. Cong. Phon. Sci., San Francisco, pp. 
611-614. 
Steven Greenberg 1999. Speaking in shorthand - A 
syllablecentric perspective for understanding 
pronunciation variation. Speech Communication, 
29:159-176.  
Hindi Bangla English – Tribhasa Abhidhaan. 2001 
Sandhya Publication  
John E. Hopcroft and Jeffery D. Ullman 1979. 
Introduction to Automata Theory, Languages 
and Computation, Addison-Wesley, USA  
Rene Kager 1999. Optimality Theory. Cambridge 
University Press 
S. Kaira 1976. Schwa-deletion in Hindi. Language 
forum (back volumes), Bhari publications, 2 (1)  
Peter F. MacNeilage and Barbara L. Davis 2000. 
On the Origin of Internal Structure of Word 
Forms. Science, 288:527-31 
B. G. Misra 1967. Historical Phonology of 
Standard Hindi: Proto Indo European to the 
present. Cornell University Ph. D. dissertation 
Manjari Ohala 1977. The Treatment of 
Phonological variation: An example from Hindi. 
Lingua, 42: 161-76 
Manjari Ohala. 1983. Aspects of Hindi Phonology, 
volume II. MLBD Series in Linguistics, Motilal 
Banarsidass, New Delhi.  
Bhuvana Narasimhan, Richard Sproat and G Kiraz. 
2001. Schwa-deletion in Hindi Text-to-Speech 
Synthesis. Workshop on Computational 
Linguistics in South Asian Languages, 21st 
SALA, Konstanz  
Martin A. Nowak, Natalia L. Komarova and Partha 
Niyogi 2002. Computational and Evolutionary 
Aspects of Language, Nature, 417:611-17 
B. R. Pray 1970. Topics in Hindi – Urdu grammar. 
Research Monograph 1, Berkeley: Center for 
South and Southeast Asia Studies, University of 
California 
Bernard Tranel 1999. Optional Schwa Deletion: on 
syllable economy in French. Formal 
Perspectives on Romance Linguistics, Ed. By J. 
Mark Authier, Barbar S. Bullock, & Lisa A. 
Reed.  
T. Vennemann 1988. Preference Laws for Syllable 
Structures. Mouton de Gruyter, Berlin 
