Pronouncing Text by Analogy 
Robert I. Damper and John EG. Eastmond 
Image, Speech and Intelligent Systems (ISIS) Research Group, 
Department of Electronics and Computer Science, 
University of Southampton, 
Southampton SO17 IBJ, 
UK. 
{ridlje}@ecs. soton, ac.uk 
Abstract 
Pronunciation-by-analogy (PbA) is an emer- 
ging technique for text-phoneme conversion 
based on a psychological model of read- 
ing aloud. This paper explores the impact 
of certain basic implementational choices 
on the performance of various PbA mod- 
els. These have been tested on their abil- 
ity to pronounce sets of short pseudowords 
previously used in similar studies, as well as 
lexical words temporarily removed from the 
dictionary. Best results of 85.7% and 67.9% 
words correct are obtained lor the pseudo- 
words and lexical words respectively, cast- 
ing doubt on certain previous-reported per- 
formance figures in the literature. 
1 Introduction 
Pronunciation-by-analogy (PbA) is an influential psy- 
chological model of the process of reading aloud. 
In PbA, most words are pronounced by retrieving their 
phonemic form from the readers's lexicon, or diction- 
ary. The pronunciation for a 'novel' word not in the 
lexicon, however, is derived not by the application of 
abstract letter-to-sound rules hut is 'assembled' from 
the (known) pronunciations of words that it resembles. 
PbA has obvious application to text-to-speech conver- 
sion by machine. 
Although PbA programs have been presented in the 
literature, they are they are few in number. Ded- 
ina and Nusbaum (1991) describe PRONOUNCE: a 
rather simple system for English. Sullivan and Damper 
(1990; 1992; 1993) describe a considerably more 
complex and developed system, but which apparently 
yields a much poorer perfornmnce. 
As a psychological theory, PbA is under-specified: 
offering little meaningfifl guidance on the implement- 
ation choices which confront the programmer. Indeed, 
Sullivan and Damper (1993) show that such choices 
can have a profound impact on performance. In this 
paper, we seek to understand how Dedina and Nus- 
baum's largely unjustified implementational choices 
affected their results and, thereby, to resolve the con- 
flict between their performance claims and Sullivan 
and Damper's. 
2 Psychological Background 
In the standard dual-route model of reading aloud 
(Coltheart, 1978), there is a lexical route for the pro- 
nunciation of known words and a parallel route util- 
ising abstract letter-to-sound rules for the pronunci- 
ation of unknown ('novel') words. Arguments for 
dual-route theory cite the ability to pronounce pseudo- 
words (non-words conforming to the spelling patterns 
of English), latency difference effects between regular 
and exception words, and apparent double dissociation 
between the two routes in dyslexia (see Humphreys 
and Evett, 1985). However, all these observations can 
arguably be explained by a single route. One pervasive 
idea is that pseudowords are pronounced by analogy 
with lexical words that they resemble (Baron, 1977; 
Brooks, 1977; Glushko, 1979; 1981; Brown and Be- 
sner, 1987). Glushko, for instance, showed that "ex- 
ception pseudowords" like tave take longer to read 
than "regular pseudowords" such as taze. Here, taze 
is considered as a "regular pseudoword" since all its 
orthographic 'neighbours' (raze, gaze, maze etc.) have 
the regular vowel pronunciation/el/. By contrast, tave 
is considered to be an "exception pseudoword" since it 
has the exception word (have,/hay/) as an orthographic 
neighbour. Thus, according to Glushko (1979), the 
"assignment of phonology to non-words is open to lex- 
ical influence". This is at variance with the notion of 
two independent routes to pronunciation. Instead: 
"it appears that words and pseudowords 
are pronounced using similar kinds of or- 
thographic and phonological knowledge: 
the pronunciation of words that share or- 
thographic features with them, and spe- 
cific spelling-to-sound rules for multiletter 
spelling patterns." 
268 
There are two tbrms o1' PbA: explicit mmlogy 
(Baron, 1977) is a conscious strategy of recalling a 
similar word and modifying its pronunciation, whereas 
in implicit analogy (Brooks, 1977) a pronunciation 
is derived from generalised phonographic knowledge 
about exisling words. The latter has obvious com- 
monalities with most single-route, conncctionist mod- 
els (e.g. Seinowski and RosenhErg, 1987) in which 
the generalised knowledge is learned (e.g. by back- 
propagation) its it set of weights, and the network has 
no holistic notion of the concept 'word'. 
Until the recent advent of computational PbA mod- 
els, analogy 'theory' could only be considered seri- 
ously underspecilied. Clearly, its operation nmst de- 
pend critically on some measure of similarity, and 
"without a metric for similarity and without a specific- 
ation of how similar is similar enough, the concept of 
analogy by similarity offers little insight" (Glushko, 
1981, p. 72). Further, as detailed by Brown and Be- 
sner (1987), the operation of IExical analogy must be 
consmfined by factors such as: 
• the size of the segment shared between novel and 
lexical word; 
• its position in the two strings; 
• its tiequency of occurrence in the hmguagc; 
• and the frequency of occurrencE of the words con- 
taining it; 
none of which had then received serious consideration. 
Accordingly, they write: "Extant analogy models are 
not capable el: predicting the ot|tconte Of assembly op- 
erations for all possiblc strings." 
In particular, the 'theory' gives no principled way 
el' deciding the orthogral~hic neighbours of it novel 
word which are deemed to intluence its pronunciation 
whereas a computational model must (spccilically or 
otherwise) do so. 
3 Existing PbA Programs 
3.1 Dedina and Nusbaum's System 
Tim overall structure el' PRONOUNCf~; is as shown 
in Fig. 1. The Icxical datalmsc consists o1' "approx- 
imately 20,000 words based on Webster's tbcket Dic- 
tionary" in which text and phonemes have been auto° 
matically aligned. Dedina and Nusbaum acknowledge 
the crude natnre of their alignment procedure, saying 
it "was carried out by a simple Lisp program that only 
uses knowledge about which phonemes are consonants 
and which are w)wels." 
An input string is matched in turn against all ortho- 
graphic entries in the lexicon. The process starts with 
the input string and the current dictionary entry left- 
aligned. Ilfformatinn about matching letter substrings 
INPUT (spelling pal-tern) 
Lexical 
Lelter-phoneme k/ alignment \] 
1 
Substring matching 
Build pronunciation 
lattice 
Decision function 
OUTPUT (pronunciation) 
Figure 1: l)cdina and Nusbaum's PRONOUNCE. 
- and their corresponding phoneme substrings in the 
dictionary entry under consideration - is entered into 
a pronunciation lattice its detailed below. The shorter 
of tile two strings is then shifted right by one letter and 
the process repeated. This continues until the two are 
right-aligned, i.e. the number of right shifts is equal to 
the difference in length between the two strings. The 
process is repeated for all words in the dictionary. 
A node of the lattice represents a matched letter, Li, 
at some position, i, in the input, as illustrated in 
Fig. 2. The node is labelled with its position index i 
and with the phoneme which corresponds to Li in the 
matched suhstring, Pim say, for the mth matched sub- 
siring. An arc is placed from node i to node j if there 
is a lnatched substring starting with Li and ending 
with L i. The arc is labelled with the phonemes in- 
termediate between l'/m and Pj,,, in tim phoneme part 
of the matched substring. Note that the empty string 
labels arcs corresponding to bigrams: the two symbols 
of the bigram label the nodes at either end. Addition- 
ally, arcs are labelled with a "frequency" count which 
is incremented by one each time that substring (with 
that pronunciation) is matched during the pass through 
the lexicon. Finally, there is a Start node at position 0 
and an End node at position one greater than the length 
of the input string. 
269 
S It E 
----___<_ 0 
A D 
/ __ d/ 
\] 
Figure 2: Partial pronunciation lattice for the pseudoword shead. 
A possible pronunciation for the input corresponds 
to a complete path through its lattice from Start to End, 
with the output string assembled by concatenating in 
order the phoneme labels on the nodes/arcs. The set of 
candidate pronunciations is then passed to the decision 
function. Two (prioritised) heuristics are used to rank 
the pronunciations, and the top-ranking candidate se- 
lected as the output. The first is based on path length. 
If one candidate corresponds to a unique shortest path 
(in terms of number of arcs) through the lattice, this is 
selected as the output. Otherwise, candidates that tie 
are ranked on the sum of their arc "frequencies". 
Dedina and Nusbaum tested PRONOUNCE on 70 
of Glushko's (1979) pseudowords, which "were four 
or five characters long and were derived from mono- 
syllabic words by changing one letter". Seven subjects 
with phonetics training were asked to read these and 
give a transcription for the first pronunciation which 
came to mind. A 'correct' pronunciation for a given 
pseudoword was considered to be one produced by any 
of the subjects. A word error rate of 9% is reported. 
3.2 Sullivan and Damper's System 
Sullivan and Damper employ a more principled align- 
ment procedure based on the Lawrence and Kaye 
(1986) algorithm. By pre-computing mappings and 
their statistics, they implemented a considerably more 
'implicit' form of PbA: there is no explicit matching 
of the input string with lexical entries. Their pronun- 
ciation lattice differs, with nodes representing junc- 
tures between symbols and arcs representing letter- 
phoneme mappings. They also examine different ways 
of numerically ranking candidates, taking into account 
probabilities estimates for the letter-phoneme map- 
pings used in the assembled pronunciation. 
Given the improved alignment and candidate- 
ranking methods, better performance than Dedina and 
Nusbaum might be expected. On the contrary, Sullivan 
and Damper's best result on the full set of 131 pseudo- 
words from Glushko (1979) (plus another 5 words - 
see section 5.1) is only 70.6% (1993, p. 449). This is 
an error rate of ahnost 30%, as compared to Dedina 
and Nusbaum's 9% on the smaller test set of size 70. 
Differences in test-set size and between British and 
American English, the transcription standards of the 
phoneticians, and the lexicons employed seem insuffi- 
cient to explain this. 
4 Re-Implementing PRONOUNCE 
Our purpose was to re-implement PRONOUNCE, as- 
sess its performance, and study the impact of vari- 
ous implementational choices on this performance. 
However, the described alignment algorithm is prob- 
lematic (see pp. 71-73 of Sullivan, 1992) and needs 
to be replaced. Rather than re-implement a flawed al- 
gorithm, we have used manually-aligned data. Since 
manual alignment generally produces a better result 
than automatic alignment, we ought to produce an 
even lower error rate than Dedina and Nusbaum's 
claimed 9%. 
The performance on lexical words (temporarily re- 
moved from the lexicon) has not previously been as- 
sessed but seems worthwhile. Arguably, 'real' words 
form a much more sensible test set for a PbA system 
than pseudowords, not least because they are multi- 
syllabic. Temporary removal from the lexicon means 
that the pronunciation must be assembled by the~ana- 
logy process rather that merely retrieved in its entirety. 
Hence, we believe it is sensible and important to test 
any PbA system in this way. 
4.1 Lexical Databases 
To examine any impact that the specific lexical data- 
base might have on performance, we have used two 
in this work: the 20,009 words of Webster's Pocket 
Dictionary and the 16,280 words of the Teacher's 
Word Book (TWB) (Thorndike and Lorge, 1944). 
In both cases, letters and phonemes have previously 
been hand-aligned for the purposes of training back- 
propagation networks. The Webster's database is that 
used by Sejnowski and Rosenberg (1987) to train and 
test NETtalk. The TWB database is that used by Mc- 
Culloch, Bedworth and Bridle (1987) for NETspeak. 
270 
The phoneme inventory is of size 52 in both cases, 
including the null phoneme but excluding stress sym- 
bols. We leave the very important problem of stress 
assignment for later study. 
4.2 Re-hnplementation Details 
The re-implementation was programmed in C on a 
Hewlett~Packard 712/80 workstation running HP-UX. 
A 'direct' version scores candidates using Dedina and 
Nusbaum's method with its two prioritised heurist- 
ics: we call this model D&N. Two other methods l'or 
scoring have also been implemented. In one, we re- 
place the second (maximum sum) heuristic with the 
maximum product of the arc frequencies: we call this 
model PROD. (It still selects primarily on the basis 
of shortest path length.) We have also inlplemented a 
version which uses a single heuristic. This takes the 
product along each possible path from Start to End 
of the mapping probabilities for that arc. These are 
computed using Method 1 (a priori version) of Sulli- 
van and Damper (1993, pp. 446-447). For all paths 
corresponding to the same pronunciation, these wdues 
are summed to give an overall score for that pronun- 
ciation. We call this the MP model. The final product 
score is not a proper probability for the assembled pro- 
nunciation, since the scores do not sum to one over all 
the candidates. 
The 'best' pronunciation is found by depth-lirst 
search of the lattice, implemented as a preorder tree 
traversal. For the D&N and PROD models, paths were 
pruned when their length exceeded the shortest \[i)und 
so far for that input, leading to a uselul reduction in run 
times. A similarly motivated pruning was carried out 
for the MP model. If any product fell below a threshold 
during traversal, its corresponding path was discarded. 
The threshold used was e times the maximum product 
score found so far, with ~ set by at 10 -3. While this 
may have led to the pruning of a path contributing to 
the 'best' pronunciation, its contribution would be very 
small. Again, this gave a very significant improvement 
in run times for the testing of lexical words (section 5.2 
below) but was unnecessary \['or the testing of pseudo- 
words. 
5 Results 
5.1 Pseudowords 
Pronunciations have been obtained lot: 
• the 70 pseudowords froln Glushko (1979) used by 
Dedina and Nusbaum to test PRONOUNCE. The 
'correct' pronunciation for these strings is taken 
to be that given by Dedina and Nusbaum (1991, 
pp. 61-62). We refer to this test set as D&N 70. 
• the lull set of 131 pseudowords from Glushko 
plus two others (goot, pome) plus two lexical 
words (cat and play) plus the pseudohomophone 
kwik, as used by Sullivan (1992). The 'correct' 
pronunciations are those read aloud by Sullivan's 
20 non-phonetician subjects, and transcribed by 
him as British Received Pronunciation. We refer 
to this test set as Sul1136. Our expectation is 
that the error rate will be relatively high for this 
test set, partly because of its larger size but more 
importantly because the subjects' dialect of Eng- 
lish is British RP rather that general American, 
i.e. there is a very significant inconsistency with 
the lexical databases. 
The output has been scored on words correct and also 
on symbol score (i.e. phonemes correct) using the 
Levenshtein (1966) string-edit distance as shown in 
Table 1. 
Our best comparison with Dedina and Nusbaum 
(D&N70 test set, D&N model, Webster's database) 
gives a figure of 77.1% words correct. This is enorm- 
ously poorer than their approximately 91% words cor- 
rect - yet the implementation, reference pronunci- 
ations and test set are (as far as we can tell) identical. 
The only relevant difference is that the Webster's data- 
base is antomatically-aligned in their work and hand- 
aligned in ours. The clear expectation, given the crude 
nature of their alignment, is that they should have ex- 
perienced a higher error rate, not a dramatically lower 
one. Overall, this result accords far more closely with 
Sullivan and Damper (1993) whose best word score for 
automatic alignment (and using smaller databases but 
a larger test set) was just over 70%. 
The re-implementation made 16 errors under the 
above conditions. Dedina and Nusbaum's claim of 9% 
words correct amounts to just 6 errors, 3 of which are 
the same as ours. The commonest problem is vowel 
substitution. It is possible to discount a very few errors 
as essentially trivial, reducing the error rate marginally 
to some 20%. We conclude, therefore, that Dedina and 
Nusbaum's reported error rate of 9% is unattainable. 
In our opinion, a major deficiency of the simple 
shortest-path length heuristic is that the output can be- 
come unreasonably sensitive to rare or unique pronun- 
ciations. For instance, mone receives the strange pro- 
nunciation /moni/by analogy with anemone. Also, 
the pseudoword shead receives the bizarre, vowel- 
less pronunciation /f___d/ (where '2 denotes the 
null phoneme) when using the D&N model and the 
TWB database. As illustrated in Fig. 2 earlier, this 
turns out to be a result of matching the unique but long 
mapping head --+/_ __d/as in forehead --+ Itbr .... d~ 
(arc li'equency 1) in conjunction with the very com- 
mon mapping sh -+/J'_/as in she and shed (arc fre- 
quency 174) which swamps the overall score of 175. 
The same bizarre pronunciation does not occur with 
the PROD model. In this case, the path through the 
271 
Table 1: Results for PbA of pseudowords with both dictionaries. See text tot further specification. 
Test set implementation 
D&N 70 D&N 
Sul1136 
PROD 
MP 
D&N 
PROD 
MP 
Webster's (%) TWB (%) 
words\[phonemes words\[phonemes 
77.1 94.3 70.0 92.6 
82.9 95.9 78.6 94.9 
85.7 96.6 80.0 95.3 
75.0 93.6 72.1 93.1 
80.1 95.0 76.5 94.5 
83.8 95.9 81.6 95.7 
(/e/, 3) node has a product score of 12 x 30 = 360 for 
the pronunciation/fed/ which considerably exceeds 
the score of 174 for/fd/. 
Replacing the arc-sum heuristic of the D&N model 
by arc-product as in the PROD model leads to 
a considerable increase in performance, e.g. from 
77. 1% words correct to 82.9% for the D&N 70 test set 
with Webster's database. In turn, the MP model per- 
forms better than PROD in all cases. 
For the Sull 136 test set, our expectation of poorer 
performance (because of the larger test set and incon- 
sistency between of dialect between the target pro- 
nunciations and the lexical databases) is borne out for 
Webster's dictionary. For TWB, however, the perform- 
ance difl'erence between test sets is less consistent. 
5.2 Lexical Words 
The primary ability of a text-to-speech system must 
be to produce correct pronunciations lbr lexical words 
(rather than pseudowords) which just happen to be ab- 
sent from the system's dictionary. Accordingly, we 
have tested the PbA implementations by removing 
each word in turn from its relevant database, and ob- 
taining a pronunciation by analogy with the remainder. 
In these tests, the transcription standard employed by 
the compilers of the dictionary becomes its own ref 
erence and problems of transcription inconsistencies 
between input strings and lexical entries are avoided. 
Results for the testing of lexical words are shown 
in Table 2. Again there are consistent performance 
differences with the 'standard' D&N model worst and 
the mapping probability (MP) mode\[ best. All models 
perform better with the TWB database than with Web- 
ster's, probably simply because of its smaller size. 
For some lexical words, no pronunciation at all was 
produced because there was no complete path from 
Start to End in the lattice. This occurred for 92 of 
the TWB words and 117 of the Webster's words irre- 
spective of the scoring model. This is a serious short- 
coming: a PbA system should always produce a best- 
attempt pronunciation, even if it cannot produce the 
correct one. Sometimes, this failure is a consequence 
of the lbrm of pronunciation lattice in which nodes are 
used to represent the 'end-points' of mappings. One 
of the inputs for which no pronunciation was found 
is anecdote, whose (partial) lattice is shown in Fig. 3. 
There is in fact no arc in the complete lattice between 
nodes (/k/, 4) and (/d/, 5) because there is no cd -+/kd/ 
mapping anywhere in either dictionary. Nor is there 
an ecd or cdo trigram - with or without the right end- 
point phonemes - which could possibly bridge the gap. 
This problem is entirely avoided with the Sullivan and 
Damper style of lattice, because the shortest-length arc 
corresponds to a single-symbol mapping rather than 
to a bigram (which may be unique). Thus, there will 
always be a 'default' single-symbol mapping corres- 
ponding to the commonest pronunciation of the let- 
ter. This is not to say that Sullivan and Damper's sys- 
tem will necessarily produce the correct output here: 
it ahnost certainly will not because of the rarity of the 
c -+/k/mapping in the _d context. 
Another input which thils to produce a pronunci- 
ation is aardvark. The problem here is not that there 
is no aa bigram in the dictionary (which is found in 
words such as bazaar), but that it only appears to- 
wards the end of other words. Dedina and Nusbaum's 
strategy ol' performing substring matching only over a 
restricted range (the number of matching comparisons 
is equal to the difference in length between the input 
string and lexical entry) is at the root of this problem. 
6 Conclusions and Discussion 
We lind that Dedina and Nusbaum's reported er- 
ror rate of 9% cannot be reproduced: our figure is 
about two or three times that. Because of the short- 
comings which emerge in this work, we believe the 
problem lies with PRONOUNCE rather than our re- 
implementation. Overall, our results are in much 
closer agreement with Sullivan and Damper's word er-- 
ror rates of almost 30% on a similar test set. 
This work suggests several useful ways in which tile 
perlk)rmance of PbA systems might be improved. Our 
best results are obtained with a scoring method based 
on a priori mapping probabilities. According to Sul- 
272 
q/rifle 2: Results for PbA of dictionary words. 
hnplcmentation 1\[ Wcbster's (%) ~WB (%) 
JLw  i w,,ra  i fi  mos 
MP I~ ')1.2 6~.9 ~ 93.5 
A N E C 
/Ol.'/ 
'Qff \'~J ki/ k!/ 
D O T E 
/ ot:~t / 
Figure 3: Simplilied prolmnciation lattice lor the lcxical word anecdote which fifils to produce any pronunciation. 
liwm and Damper (1993), a posteriori mapping prob- 
abilities may do evcn better. Also, the type of pronun- 
ciation lattice used by Sullivan and Damper, in which 
nodes correspond to thejuncturcs between symbols, is 
likely to be superior. The impacl of different align- 
ment strategies should repay study. Finally, we intend 
to assess the impact of incorl)orating inlormation about 
word frequency in the analogy process. 
Acknowledgement 
This work was funded by the UK Economic and So- 
cial Research Council via rescarch grant R000235487: 
"Speech Synthesis by Analogy". 
References 
Baron, J. (1977). Mechanisms for pronouncilag printed 
words: use and acquisition. In Basic l'rocesses in Reading: 
Perception attd (7omprehension (1). l~aBcrgc and S. Samuels, 
eds.), pp. 175--216. l,awrence l';rlbamu, tlillsdalc, NJ. 
Brooks, L. (1977). Non-analytic correspondences and pat-. 
tern in word pronunciation. In Attention and Petjbrm- 
ance VII (J. Rcnquin, ed.), pp. 163-177. 1 ,awrcnce I';rlbaum, 
Hillsdalc, NJ. 
Brown, P. and Besner, I). (1987). The asseml)ly of phon- 
ology in oral reading: a new model. In Attention and Per- 
.formatwe Xlh the Psychology oJ'Reading (M. Colthcart, ed.), 
pp. 471-489. Lawrence Edbatnn, London. 
Coltheart, M. (1978). Lexical access in simple reading 
tasks. In Strategies of h!/brmation l'rocessing (G. tJnder- 
wood, ed.), pp. 151-.216. Academic, I xmdon. 
I)edina, M.J. and Nusbaum, I I.C. (1991). I)I,IONOIJNCE: 
a program for l)ronunciation by analogy. Compuler Speech 
and lzmguage, 5, 55--64. 
Glushko, R.J. (1979). The organization and actiw~tion of or- 
thographic knowledge in reading aloud. ,hmrna\] of l'Speri~ 
mental Psychology: ltuman l>erctT)tion t#ld I'e~jbrmance, 5, 
674-691. 
Glushko, R.J. (1981). Principles for pronouncing print: the 
psychology of phonography. In hzteractive Processes in 
Reading (A.M. Lesgold and C.A. Perlbtti, eds.), pp. 61-84. 
Lawrence Erlbaum, ttillsdale, NJ. 
llumphreys, G.W. and Evett, I,.J. (1985). Are there inde- 
pendent lexical and nonlexical routes in word processing? 
An cwlluation of the dual-route theory of reading. Behavi- 
oral and lh'ain Sciences, 8, 689-740. 
\[,awrence, S.G,C. and Kaye, G. (1986). Alignment of 
phonemes with their corresponding orthography. Computer 
Speech and Language, 1, 153-165. 
Levenshtein, V.I. (1966). Binary codes capable of corrccting 
deletions, insertions and reversals. CybetTtetics and Control 
Theory, 10, 707-710. 
McCutloch, N., Bedworth, M. and Bridle, J.S. (1987). NET- 
speak - a re4mplementation of NETtalk. Computer Speech 
and lxmguage, 2, 289-301. 
Sejnowski, T.J. and P, osenberg, C.R. 0987). Parallel net- 
works that loam to pronounce English text. Conwlex Sys- 
tems, 1, 145-152. 
Sullivan, K.RH. (1992). Synthesis-by-Analogy: a Psycho- 
logically Motiwtted Approach to Textqo-Speech Convelxion, 
Phi) rl'hcsis, Department of Electronics and Computer Scio 
ence, University of Southampton, UK. 
Sullivan, K.P.II. and l)amper, R.I. (1990). A psychologic- 
ally governed approach to novel-word pronunciation within 
a text-to.-spcech system, l'roceedings oflEl~E International 
Conj~;mnce on Acoustics, Speech and Signal Processing 
( ICASSt' '90), Vol. I, Albuquerque, NM, pp. 341--344. 
Sulliwm, K.RII. and I)ampcr, R.I. (1992). Novel-word pro- 
nunciation within a text-to-speech system. In rlMking Ma- 
chines: 7beeries, Models atul Applications (G. Bailly and C. 
Bcno'fl, cds.), pp. 183--195. Elsevier (Northqlolland), Ams- 
terdanL 
Sullivan, K.P.H. and Damper, R.I. (1993). Novel-word pro-. 
nunciation: a cross-lallgtlage study. Speech (;ontmunicatiotz, 
13, 44l. 452. 
Thorn,like, E.L. and l,orgc, 1. (1944). The Teachetw' Word 
Book oJ30,O00 Words. Teachers' College, Columbia Uni- 
versity, NY. 
273 
