TWO SIMPLE PREDICTION ALGORITHMS 
TO FACILITATE TEXT PRODUCTION 
Lois Boggess 
P.O. Drawer CS 
Mississippi State University 
Mississippi State, MS 39762 
ABSTRACT 
Several simple prediction schemes are 
presented for systems intended to facili- 
tate text production for handicapped 
individuals. The schemes are based on 
single-subject language models, where the 
system is self-adapting to the past 
language use of the subject. Sentence 
position, the immediately preceding one 
or two words, and initial letters of the 
desired word are cues which may be used 
by the systems. 
INTRODUCTION 
For some years we have been investi- 
gating the use of a sizeable sample of a 
particular individual's language habits 
in predicting future language use for 
that individual. The research has taken 
two directions. 
One of these, the HWYE (Hear What 
You Expect) system, builds a large lan- 
guage model of the past language history 
of the individual, with special emphasis 
on the most frequent words of that 
person, and the result is used in speech 
recognition. In studying the language 
model developed by the HWYE system, 
several simple predictive schemes were 
noted which are capable of anticipating, 
during the generation of a sentence, a 
small set of words from which the next 
desired word can be selected. The two 
schemes described here are used for text 
generation (not speech recognition) in a 
format that could be of use to a physi- 
cally handicapped person; hence the 
schemes have no right context available. 
One of the schemes does use left context, 
and the other uses only sentence position 
as "context'. Both are implemented on 
IBM-PC systems with minimal memory 
requirements. 
MOTIVATION 
One hundred English words account for 
47 per cent of the Brown corpus (about 
one million words of American English 
text taken from a wide range of sources). 
It seems reasonable to suppose that a 
single individual might in fact require 
fewer words to account for a large 
proportion of generated text. From our 
work on the HWYE system it was known 
that 75 words accounted for half of all 
the text of Vanity Fair, a 300,000 word 
Victorian English novel by Thackeray 
(which incorporated a fairly involved 
syntax, much embedded quotation, and 
passages in dialect and in French) 
\[English and Boggess, 1986\]. We further 
found that 50 words accounted for half of 
all the verbiage in a 20,000 word set of 
sentences provided by an individual who 
collaborated with us. This latter 
corpus, called the Sherri data, is a set 
of texts provided by a speech-handicapped 
individual who uses a typewriter to 
communicate, even with her family; it is 
conversational in nature, as can be seen 
in Figure 1. Most of the work reported 
in this paper gives special attention to 
the set of words required to account for 
half of all the verbiage of a given 
individual. We refer to this set as the 
set of high-frequency words. 
33 
You said something about a magazine that <namel> had 
about computers that I might like to borrow. 
I would some time. 
I think we have to pick up the children while <name2> 
is in the hospital. 
I want to visit her in the hospital. 
But you have to lift me up to the window for me to see 
the baby. 
Well, it's May first now. Help! 
I thought it would not be so busy but it looks like it 
might be now. 
Figure 1. Sample set of contiguous sentences in Sherri data 
It seems reasonable to suppose that 
for conversational English, approximately 
50 words may account for half of the 
verbiage of most English users. From 
the standpoint of human factors, an 
argument could be made that one should 
simply put the 50 words up on the screen 
with the alphabet and thus be assured 
that half of all the words desired by the 
user were instantly available, in known 
locations that the user would quickly 
become accustomed to. Constantly 
changing menus introduce an element of 
user fatigue \[Gibler and Childress, 
1982\]. That argument may especially make 
sense as larger screens with more lines 
per screen and more characters per line 
become more common. 
If we limit ourselves to the top 20 
most frequent words as a constant menu, 
only about 30 per cent of the user's 
verbiage is accounted for. However, it 
was observed, while working with the 
HWYE system, that if one looked at the 
top 20 words for any given sentence 
position, one did not see the same set of 
words occurring. Clearly the high 
frequency words (the set that comprise 
half of word use) are mildly sensitive to 
"context" even when "context" is so 
broadly defined as sentence position. 
Different subsets of the 50 member set of 
high frequency words appear in the set of 
20 most frequent words for a given 
sentence position. Moreover, after 
processing approximately 2000 sentences 
from the user, it was still the case that 
some of the top 20 words for a given 
position were not members of the high 
frequency set at all. For example, the 
word "they', a member of the menu for the 
first sentence position Isee Figure 2) 
and hence one of the 20 most frequent 
words to start a sentence, is not a 
member of the global high frequency set. 
A preliminary analysis by English 
suggested that, whereas a constant 
"prediction" of the top 20 most frequent 
words would yield a success rate of 30 
per cent, predicting the top 20 most 
frequent words per position in sentence 
would yield a success rate of 40 per 
cent. 
~CONTEX7" AS SENTENCE POSITION 
The simplest scheme, which has been 
built as a prototype on an IBM PC with 
two floppy disk drives, presents the user 
with the top 20 most frequent words that 
the user has employed at whatever 
position in a sentence is current. For 
example, Figure 2 shows the screen 
presented to the user at the beginning of 
production of a sentence. On the left is 
a list of the 20 words which that 
particular user is known to have used 
most often to begin sentences. On the 
right is the alphabet, which is normally 
available to the user; and in other 
places on the screen are special 
functions. (Selection of words, letters 
34 
1 but 
2 oan 
3 oould 
4 do 
S he 
6 hot* 
? I 
8 I°N 
9 if 
10 it 
II it) s 
12 Lois 
13 she 
14 that 
15 the 
16 they 
17 we 
18 what 
19 when 
2e uou 
SPELL 
CAPITAL 
• b o 
PUNCTUATION 
HELP-NENU S h I 
ENDING 
m n o 
NUNDER 
SPECIAL • t u 
REUlEN 
U z 
HARD-COPY 
SAUE-SENT 
j k 1 
p q r 
v w x 
HEN ERASE qUIT 
HEN SENTENCE: 
Figure 2. Initial Screen 
and functions is made by mouse, though 
the actual selection mechanism is 
separated from the bulk of the code so 
that replacement with another selection 
mechanism should be relatively easy to 
implement.) The sentence is built at the 
bottom of the screen. If the user 
selects a word from the menu at the left, 
it is placed in first position in the 
sentence, and a second menu, consisting 
of the 20 most frequent words that the 
user has used in second place in a 
sentence, appears in the left portion of 
the screen. After a second word has been 
produced and added to the sentence, a 
third menu, consisting of the 20 most 
frequent words for that user in third 
place in a sentence, is offered, and so 
on. 
At any time the user may reject the 
lefthand menu by selecting a letter of the 
alphabet. Figure 3 shows the screen after 
the user has produced two words of a 
sentence and has begun to spell a third 
word by selecting the letter "a +. At this 
point, the top 20 most frequently used 
words beginning with +a" have been offered 
at the left. If the desired word is not 
in the list, the user continues by select- 
ing the second letter of the desired word 
(in this case, "n'). The left-hand menu 
becomes the 20 most frequently used words 
beginning with the pair of letters given 
so far. As is shown in Figure 4, there 
are times when fewer than 20 words of a 
given two-letter starting combination have 
been encountered from the user's past 
history, in which case this algorithm 
offers a shortened list. 
In the case illustrated, the desired 
word was on the list. If it were not, the 
user would have had to spell out the en- 
tire word, and it would have been entered 
into the sentence. In either case, the 
system subsequently returns to offering 
the menu of most-frequently-used words for 
the fourth position, and continues in 
similar fashion to the end of the 
sentence. 
L• 
2 able 
3 •bout 
4 •£ter 
5 afternoon 
6 apaln 
7 all 
8 am 
9 an 
le and 
12 apple 
13 April 
14 are 
15 a~ound 
16 as 
17 ask 
18 asked 
19 at 
2e •unt 
• b o d • F 
g h i J k 1 
• t u v w )< 
NEW SENTENCE: 
I have 
Figure 3: 
1 animal 
2 animals 
3 Anita 
4 anniversary 
S AnnM 
6 another 
? &nsuer 
8 answer• 
9 any 
IO •nvone 
II •npth|n• 
User has selected "a" 
• b o d • £ 
• k I J k I 
• t u v ~ x 
NEN SENTENCE: 
I have 
Figure 4: User has selected "a-n" 
35 
The system keeps up with how often a 
word has been used and with how many 
times it has occurred in each position in 
a sentence, so that from time to time a 
word is promoted to one of the top 20 
alphabetic or top 20 position-related sets 
of words. For details on the file organi- 
zation scheme that allows this to be done 
in real time, see Wei \[1987\]. Details on 
the mouse-based implementation for IBM 
PC's are available in Chow \[1986\]. 
A SECOND ALGORITHM 
An alternative predictive algorithm 
has been implemented which replaces the 
sentence-position-based first menu. It 
pays special attention to the 50 most 
frequently used words in the individual's 
vocabulary (the high-frequency words) and 
to the words most likely to follow them. 
By virtue of their frequency, these are 
precisely the words about which the most 
is known, with the greatest confidence, 
after a relatively small body of input 
such as a few thousand sentences. 
For each of the 50 high-frequency 
words, a list is kept of the top 20 most 
frequent words to follow that word. Let 
us call these the first order followers. 
For each of the first order followers, 
there is a list of second-order followers: 
words known to have followed the two 
word sequence consisting of the high- 
frequency word and its first order 
follower. 
For example, the word "I" is a high- 
frequency word. The first order followers 
for "I" include the word "wol)ld'. The 
second-order followers for "I would" 
include the word "like'. (See Figure 5.) 
The second-order followers for "I would" 
also include many one-time-only followers, 
as well, so the system maintains a 
threshold for the number of oceurrances 
below which a word is not included in the 
list of second-order followers. The 
reasoning is that a word's having occurred 
only once in an environment that by 
definition occurs frequently may be taken 
as counter-evidence that the word should 
be predicted. 
Rather than predict a word with low 
reliability, one of two alternatives are 
taken. If the first-order follower is 
itself a high-frequency word, then low- 
reliability second-order followers may be 
replaced with the first-order follower's 
own followers. ('Would" is a first-order 
I o 
Figure 5. 
..~-! thi,k ,-~--'" 
don't *,-~, 1 
hope ~. ! ' 
i was 
wish 
like 
will 
have 
want 
wonder 
got 
r -, ~z 
I'll 
the 
we 
it 
It'S 
oF 
Vou 
really 
wan t 
have 
,,. ......... Q 
First- and second- followers 
for "I" 
follower of "I" and is itself a high- 
frequency word. There are relatively few 
reliable second-order followers to "would" 
in the left context of "I', so the list is 
augmented with first-order followers of 
"would" to round out a list of 20 words.) 
The other alternative, taken when the 
first-order follower is not a high- 
frequency word, is to fill out any short 
list of second-order words with the high- 
frequency words themselves. 
This algorithm is related to, but 
takes less memory and is less powerful 
than a full-blown second order Markov 
model. Each state in a second-order 
(trigram) Marker model is uniquely 
determined by the previous two inputs. 
For an input vocabulary of 2000 words, the 
number of mathematically possible states 
in a trigram Marker model is 4,000,000, 
with more than 8 billion arcs intercon- 
necting the states. Fortunately, in the 
real world most of these mathematically 
possible states and arcs do not actually 
occur, but a trigram model for the real 
world possibilities is still quite large. 
We experimented with abstracting the 
input vocabulary by restricting it to the 
50 highest-frequency words plus the 
pseudo-input OTHER onto which all other 
words were mapped. When we did so, the 
number of states and arcs in the various 
order Markov models was still fairly large 
for the real world data \[English and 
Boggess, 1986\]. As Figure 6 shows, for 
example, the rate of growth for a fourth- 
order abstract Markov model (just the 50 
highest-frequency words plus OTHER plus 
end-of-sentence) is in the neighborhood of 
250 new states and 450 new arcs per 1000 
36 
Sherri data Thackeray data 
words new states new arcs new states new arcs 
1000 527 677 639 830 
2000 469 620 624 818 
3000 471 636 476 705 
4000 399 562 467 716 
5000 397 566 463 714 
6000 391 579 437 668 
7000 337 507 389 642 
8000 311 476 370 628 
9000 323 500 361 612 
10000 285 486 384 629 
11000 329 518 348 601 
12000 278 448 331 588 
13000 276 445 310 543 
14000 240 408 291 530 
15000 248 425 287 529 
16000 244 420 290 533 
17000 243 414 269 497 
18000 259 446 234 468 
Figure 6. Growth of abstracted fourth-order Marker models 
new words of text, after 17000 words of 
input. This was true for both the Sherri 
data (conversational English) and the more 
formal Thackeray data. Moreover, the 
fourth-order Marker model for the 
abstracted Thackeray data continued to 
grow. After 100,000 words of input, with 
a model of approximately 22,000 states and 
approximately 45,000 arcs, the rate of 
growth was still more than 1,000 states 
and 3,000 ares per 10,000 words of input. 
For this particular implementation, 
however, neither r. full-blown Markov 
model using total vocabulary nor an 
abstract model using the 50-word vocabu- 
lary seemed appropriate. On the one hand, 
models of the entire vocabulary confirmed 
that many multiple word sequences did 
occur regularly. Nevertheless, for any 
but the simplest order Marker models 
(orders zero and one), the vast bulk of 
the networks were taken by word combina- 
tions that occurred only once. On the 
other hand, restricting the predictive 
mechanism to only the high-frequency words 
obviously left out some of the regularly 
occurring word combinations. Our first- 
and second-follower algorithm described on 
the previous pages allows lower frequency 
words to be predicted when they occur 
regularly in combination with high- 
frequency words. 
PREDICTIVE CAPABILITIES 
The data used to test the predictive 
capabilities of the system were type- 
scripts provided by the user, who was 
utilizing a manual typewriter; it follows 
that the results were not biased by the 
user's favoring sentence patterns that the 
system itself provided. The system had 
bccn given 1750 prior scntcnces produced 
by the user and the data collected were 
for the performance of the system over the 
next 97 sentences. The 1750 sentences 
were 14,669 words in length with a vocabu- 
lary of 1512 words. Twelve sentences of 
the 1750 were a single word in length 
{e.g. "yeah", "no" and "gesundheit") and 
51 were of length 20 or greater. Average 
length of sentence for the initial body 
was 8.4 words per sentence. The first 200 
sentences included transcriptions of oral 
sentences, which were much shorter on 
average, since the user is speech handi- 
capped. If the first 200 sentences are 
omitted, the average sentence length is 
8.6 for the following 1550 sentences. 
Of the next 97 sentences generated, 
the shortest sentence was "Thanks again." 
The longest was "You said something about 
a magazine that Jenni had about computers 
that I might like to borrow." The 97 
sentences consisted of 884 words (six of 
which were numbers in digital form), for 
an average length of 9.1 words per 
sentence. 
37 
Of the 884 words, 350 were presented 
on the first menu, 373 were presented on 
the second menu (after one letter had been 
spelled), 109 were presented on the third 
menu (after two letters had been spelled),. 
2 were presented on the fourth menu (after 
three letters had been spelled, 43 were 
spelled out in their entirety, and 7 were 
numbers in digital form, produced using 
the number screen of the system. 
From the above, it is obvious that 
the device of predicting the 20 most 
frequent words by sentence position is 
successful 39.6 per cent of the time; 
42.2 per cent of the time, the desired 
word is among the 20 most frequent words 
of a given initial letter but not in the 
20 most frequent words by position; 
combining these two facts, we see that 
81.8 per cent of the time, this simple 
prediction scheme presents the desired 
word on a first or second selection. The 
desired word is offered in the first, 
second, or third menu 94.1 per cent of the 
time, and most of the rest of the time 
(5.7 per cent of total), the desired word 
is unknown to the system and is "spelled 
out', where "spelling" includes producing 
numbers. 
Although the fourth menu, consisting 
of words with a three-letter initial 
sequence, presently has a low success 
rate, it is precisely this category that 
we expect to see improve as more of the 
user's words become known to the system 
through spelling. That is, as time 
passes, we expect the user to have to 
resort to complete spelling less and less 
because the known vocabulary will include 
more and more of the actual vocabulary of 
the user. Many of the new words will be 
low frequency words that we would expect 
to find on the menu for three-letter com- 
binations after they are known. 
The second algorithm, using first- and 
second-followers of the high-frequency 
words, was run on i00 sentences, the 
shortest of which was "Help!" (94 of the 
97 test sentences for the first algorithm 
were represented in the test set for the 
second.) There were 895 words in the 
sample, of which 448 were presented on the 
first menu, 280 were presented on the 
second (after one letter had been spelled 
out, 83 on the third (after two letters 
were spelled), 1 on the fourth, and 83 
were spelled out in their entirety (this 
category included numbers). 
Running the second test gave us a 
very quick appreciation for the value of 
adding new words to the system as they 
are encountered, since this implementation 
of the second algorithm did not. One 
especially striking example was a word 
beginning with "w-o" which had never been 
used before, but which occurred five times 
in the 100 test sentences and had to be 
spelled out each time. This was especial- 
ly irritating since the "w-o" menu (third 
menu) had fewer than 20 entries and would 
have accommodated the new word. A com- 
parison of the two columns of Figure 7 
suggests that for the text held in common 
by the two tests, approximately 30 words 
had to be spelled out by the second algo- 
rithm, which were selected by menu in the 
first algorithm because it added new words 
to its data sets as they were encountered. 
PROPOSED EXTENSIONS 
We have several plans for the future, 
most of them involving the second algo- 
rithm. Our first task is to increase the 
number of sentences in the Sherri data to 
3000 and determine how much (if at all) 
an enlarged base of experience improves 
the ability of the algorithm to predict 
Sentence position algorithm 
number sentences: 97 
number of words: 884 
frequent word/left context algorithm 
number sentences: 100 
number of words: 895 
words % total 
first menu: 350 39.6% 39.6% 
second menu: 373 42.2% 81.8% 
third menu: 109 12.3% 94.1% 
fourth menu: 2 0.2% 94.3% 
spelled: 43 4.8% 99.2% 
numbers: 7 0.8% 100% 
words % total 
first menu: 448 50% 50% 
second menu: 280 31.3% 81.3% 
third menu: 83 9.3% 90.6% 
fourth menu: 1 0.1% 90.7% 
"spelled': 83 9.3% 100% 
Figure 7. Comparison of the predictive capabilities. 
38 
the desired word on the first try. 
In its present form, the system is 
reliable in its predictions after several 
hundred sentences by the user have been 
processed. We intend to take something 
like the Brown corpus for American 
English and from it create a vanilla- 
flavored predictor as a start-up version 
for a new user, with facilities built in 
to have the user's own language patterns 
gradually outweigh the Brown corpus 
initialization as they are input. 
Eventually the Brown corpus would have 
essentially no effect, or at least no 
effect overriding the user's individual 
use of language (it might serve as a 
basic dictionary for text vocabulary not 
yet seen from the user). 
We intend to investigate what effect 
generating sentences while using the 
system has on our collaborator. To date, 
she has obligingly been willing to 
continue to use a typewriter to generate 
text, but she does own a personal computer 
and is able to use a mouse. Our own 
experience in entering her sentences on 
the system has made it clear that in many 
instances she would have expressed the 
same ideas more rapidly on the system with 
a slight change in wording. Since the 
preferred words and patterns are derived 
by the system from her own language 
history, they should feel normal and 
natural to her and could influence her to 
modify her intentions in generating a 
sentence. On the other hand, a different 
handicapped individual (a quadriplegic) 
has informed us that ease of mechanical 
production of a sentence has little or no 
effect on his choice of words, and that 
would appear to be the case for our 
collaborator while she uses the 
typewriter. 
Finally, we wish to make use of the 
much larger amounts of memory available 
on personal computers by taking account of 
the followers for many of the moderate- 
frequency words. For example, in the 
sentence "would you be able..." the word 
"able" is not high frequency. Neverthe- 
less, the system could easily deduce what 
following word to expect, since every 
known occurrence of "able" is followed by 
"to'. As it happens, "to" is one of the 
top 20 most frequent words and hence 
fortuitously is on the default menu after 
the non-high-frequency word "able', but 
there are many other examples where the 
system is not so lucky. For instance, 
"pick" is usually followed by "up" in the 
Sherri data, but "pick" is low frequency 
and "up" is not on the default first menu. 
Similarly, "think" is a high-frequency 
word and has a well developed set of 
followers. "Thinks" and "thought" are not 
high-frequency and hence are followed by 
the default first menu. Yet virtually 
every follower for "thinks" and "thought" 
in the Sherri data happens to belong to 
the set of followers for "think'. We 
believe that by storing information on 
moderate frequency words with strongly 
associated followers and on clusters of 
verb forms we may significantly improve 
the success of the first menu. 
RELATED WORK 
That a small number of words account 
for a large proportion of the total ver- 
biage in conversation has been known for 
some time \[Kucera and Francis, 1967\]. 
The idea of using the first several 
letters typed by a handicapped individual 
to anticipate the next desired word has 
been used in numerous systems (e.g., 
\[Giblet and Childress, 1982\], \[Picketing 
et al., 1984\]). The Gibler and Childress 
system is typical in that it uses a few- 
thousand-word vocabulary drawn from the 
general public, plus a few hundred words 
specific to the user of the system. The 
user must type the first two letters 
before the system provides a menu of 
words beginning with the letter pair. If 
the desired word was not on the menu, the 
user had to spell the word out. It was 
felt that one letter was not informative 
enough to warrant a menu. Furthermore, 
Gilbler and Childress showed that increas- 
ing the system vocabulary degraded the 
performance of their system and they 
recommended limitation of the vocabulary 
for human factors reasons. 
By contrast, our system costs the 
user no more effort in terms of selecting 
the first two letters - if indeed they 
have needed to go that far; 80 per cent 
of the time, they haven't needed to pro- 
vide two letters. Further, there is no 
question that for our system, allowing the 
vocabulary to grow is of benefit both to 
system performance and to user satis- 
faction. 
Galliers \[1987\] describes a different 
approach for physically handicapped 
39 
persons conversant in the Bliss communi- 
cations system. Communication with Bliss 
involves a high degree of interpretation 
by the "listener', and Galliers reports an 
impressive 75 per cent success rate in 
automating such interpretation. The 
Galliers system is single-subject, as ours 
is, and it does use past history to 
facilitate interpretation. It was, how- 
ever, limited to a very small domain for 
the experiment described. 
One statistic cited by this last paper 
was that the same text produced from the 
Bliss communication, had it been produced 
by typing into a word processing system, 
would have required three times as many 
key-press operations. Our own ratio of 
key-press operations to characters 
produced was 45 per cent for the sentence 
position algorithm. That is, on average 
it took 45 presses of a mouse button to 
produce 100 characters. Part of the 
reason for such a high ratio has to do 
with punctuation, capitalization, and 
special screens such as the number screen, 
which requires not only the same number of 
presses of the button as there are digits, 
for example, but additional presses of the 
button to summon the screen and quit the 
menu. But primarily the ratio seems to 
derive from the fact that many of the 
words in any text are short - "a', "to', 
"the', "of', "in', and "on" being examples 
from this very paragraph. If the first 
menu does not contain a desired two-letter 
word, one has to spell the first letter 
and then make a selection from the second 
menu - requiring two presses of a button. 
By contrast, Bliss users commonly use 
a telegraphic style of communication and 
omit function words altogether. 
CONCLUSION 
In summary, evidence exists that for 
a system built around a single user's 
language, a prediction scheme that simply 
anticipated fifty or so words would on 
average be correct about half the time. 
Limiting such a system to only the top 20 
most frequent words would give a success 
rate of about 30 per cent. However, not 
all of the high frequency words are dis- 
tributed evenly by sentence position. A 
system that offers the top 20 most fre- 
quently occurring words for each position 
of a sentence was successful about 40 per 
cent of the time on the next 97 sentences. 
Allowing a user to reject the first set of 
words by giving the first letter of the 
desired word and offering the 20 most 
frequent words beginning with that letter 
resulted in success for the combined first 
and second menus 82 per cent of the time. 
After a training body of 1750 sen- 
tences (14,669 words), with a vocabulary 
of 1512 words, it was still the case that 
about six per cent of the desired words 
were unknown to the system. 
An alternative algorithm for the first 
offering of 20 words, based primarily on 
the right hand contexts of the high fre- 
quency words, is successful on the first 
guess 50 per cent of the time. 

REFERENCES 

Boggess, Lois and Thomas M. English, The 
HWYE speech recognition system: a user- 
specific model for expectation-based 
recognition, in Proceedings of the 25th 
Southeast Regional Conference of the 
ACM, Birmingham, 1987. 

Chow, C. L. A mouse-driven menu-based 
text prosthesis for the speech 
handicapped, M.C.S. project report, 
Mississippi State University, 1986. 

English, T. M. and Lois Boggess, A grammatical approach to reducing the statistical sparsity of language models in natural domains, Proceedings of the International Conference on Acoustics, Speech, 
and Signal Processing, Tokyo, 1986. 

Galliers, Julia, AI for special needs - 
an "intelligent" communication aid for 
Bliss users, Applied Artificial 
Intelligence, 1(1):77-86, 1987. 

Gibler, D. C. and D. S. Childress, Lan- 
guage anticipation with a computer based 
scanning aid, Proceedings of the IEEE 
Computer Workshop on Computers 
to Aid the Handicavoed, 1982. 

Kucera, H. and W. N. Francis, Computa- 
tional analysis of present-day American 
English. Brown University Press, 1967. 

Pickering, J., J. L. Arnott, J. G. Wolff, 
and A. L. Swiffin, Prediction and adap- 
tation in a communication aid for the 
disabled, Proceedings of the IFIP 
Conference on Human-Computer 
Interaction, London, 1984. 

Wei, Jan-Soong, File organization of 
Sherri System, M.C.S. project report, 
Mississippi State University, 1987. 
