Applications of~a Com~uter System 
for Transformational Grammar* 
by 
Joyce Friedman 
The University of Michigan 
Ann Arbor, Michigan, U.S.A. 
Writing a transformational grammar for even a fragment of a 
natural language is a task of a high order of complexity. Not only 
must the individual rules of the grammar perform as intended in 
isolation, but the rules must work correctly together in order to 
pro~nce the desired results. The details of grammar-writing are 
likely to be regarded as secondary by the linguist, who is most 
concerned with what is in the language and how it is generated, and 
would generally prefer to pay less attention to formal and notational 
detail. It is thus natural to ask if a computer can be used to assist 
the linguist in developing a grammar. The model is formal; there 
are a large number of details to be worked out; the rules interact 
with one another in ways which may not be foreseen. Most of the 
errors which occur in writing grammars can be corrected if only they 
are brought to the attention of the linguist. Those which cannot 
be so corrected should be of even greater interest to the writer of 
the grsmmar. 
This research was supported in part by the United States Air 
Force Electronic Systems Division under Contract F19628-C-00353 and 
by the National Science Foundation under Grant GS-2271. 
A computer system which attempts to provide this assistance 
to the writer of a grammar is now available at both the University 
1 of Michigan and Stanford University. The system is written in 
Fortran IV and runs on the II~ 360/67 computer. 
The linguistic model embodied by the system is the theory of 
transformational grammar, roughly as described by Chomsky in 
of the Theory of Syntax. \[2\] The programs represent the linguistic 
metatheory. Grammars are accepted by the programs as data. The 
program handles all three components of a transformational grammar: 
phrase structure, lexicon, and transformations. It carries out the 
full process of sentence generationj including phrase structure 
generation, lexical insertion, and transformation. 
The technical details of the particular model of transformational 
grammar have been described elsewhere \[3\]. This presentation will 
emphasize the ways in which the programs can be used, and will 
describe experiences in using them both in grammar writing and in 
teaching. 
An example of a grammar 
The notation for grammars and the use of the programs will not 
be described formally here, but will be illustrated by an extended 
example. The example consists of a small grarmnar and a sample deri- 
vation. Each part will be presented twice, first as initially 
iThis system was designed and programmed by the author 3 with 
T. H. Bredt, E. W. Doran~ T. S. Martner, and B.W. Pollack. 
prepared by linguists at the University of Montreal \[i\], and then 
as redone in the computer system. The grammar has been greatly 
reduced by selecting only those transformations which were used in 
the derivation of the sample sentence selected. 
In Figures i and 2 the phrase-structure rules are given in 
parallel, first as written by the linguists, secondly as prepared 
for input to the computer system. The computer form can be seen to 
be a linearization of the usual form, with both parentheses and 
curly brackets represented by parentheses. No ambiguity arises from 
this since the presence of a comma distinguishes the choices from 
the options. The only other differences are minor: the symbol 
"~" has been replaced by "DELTA", the sentence symbol "P" has been 
translated into English "S", and accents have been omitted. None of 
these changes is of any consequence. 
Figure 5 is a listing of a-partial lexicon. This component is 
present only implicitly in the original French gran~nar, where the 
complex symbols are indicated in the base trees. The lexicon specifies 
first the features which are used in the grammar. The list of cate- 
gory features also determines the order in which lexical insertion 
takes place. The inherent features include some which are included 
in complex symbol s in the lexicon, and others which are added only 
by transformations. Contextual features are defined in the lexicon 
by giving a structural analysis of the context in which the item can 
appear. The location in which the word may be inserted is indicated 
by an underline symbol. Thus, common nouns are marked +NCOM, which 
means that they can occur only following a determiner (DET) in a 
P 
PRE 
NEG 
PRED 
ADV 
INST 
SV 
SA 
COP 
SN 
COMPL 
DET 
DEF 
ANAPH 
DEM 
CARD 
Figure i 
Phrase Structure Rules, from \[i\] 
--, # ( PRE ) SN PRED # 
_, (i~) (~m~) 
ne pas 
INST 
v (co~) 
cop ADJ (CO~L) 
_, est 
ICDE~) N 
( | quel| ) (CARD) 
-" I~HI 
tDEM J 
-~ I cele ( la ) 
-~ ce ( i~ 
~PLURJ 
"MONTREAL FRENCH" 
PHRASES LRUCTURE 
$ = # (PRE) SN PRED # . 
PRE : (IN/) (NEG). 
NE~ : NE PAS. 
PRED : (SV (ADVINS),SA). 
ADVI~S = PAR (SN,DEL~A). 
SV = V (COMPL). 
SA = COP AD,J (CONPL). 
COP = EST. 
SN : ((SN) S, (DET) N). 
COMPL : (SN,5) (SAD. 
DET : ((DEF,QUEL))(CARD). 
DEF : (ANAPH,DEM). 
ANAPH = (CE ((CI,LA)),LE). 
DEM = CE ((CI,LA)). 
CARD = (SING,PLUR). 
SING : UN. 
PLUR = (PROCARD,QUELQUES,DEUX,IROIS). 
PROCARD : NO~SRE DE. 
$ENDPSG 
Figure 2 
Phrase Structure Rules 
Figure 3 
Lexicon 
^ 
A 
Z 
~ ^ 
A 
O V 
Z~J 
O9 V 
r~ ~ Q_ ~,~ ~' n 
---j m-- ~-- t~.l ^ I=.I I.~I E 
• "'. ~I 0.-3~ A E O 
<=I:L z Q" _-3 IQ.. -- L} 
0 "~' I-,, 
I-, O <1: l.,.. =~ V 
O {'1~ E I.~ o z {)~ :3I. U'I 
L,L.\] "--~ ~ ~ CY') V V ~" n ~r" ~"~ Q..--I V 
~'--I "-q ~ II 
rE:?: {'--" "--~ 
0 '~,.~ L=J ~-' C.) ''~ 
"--~ '~E "~" 0 
.J 
L~ 
L~L,J + 
// 
JJ (:30 . - 
~'Z'JE (3(3 / / 
(:30 Z:~-:E_ ~- 
(3(3/++OO 
~" Z :.~E (30 
,?~ rE :=~- ...l .J 0 
,,-I .-J I 0 --~ :3 D- 
IL l:I. rE --I.-I J 
i i ::::3.'E_ EI~. r~. 
--I L~J tTJ I t .'E 
LI. l.~ I.~ L~I I--, "--1 
LLI::3r'~ I t ,~:: i / 
~/} O,) I=. -E ~,LI .,~ 0 
• '~-,~ -F + l*J..r~ r~ I r-. 0 
0 0 r~r:- FE i-" r', n L~ -- OQ.. - 0 
I I O0~/'ll--,l-~: ~ I ~'~I "-~ 
{~ ~ I-, I-., -I.=J I~1 I~: /' {/~ / :3, 
:Y'E I I 0..D'} ~'f}l-, l:~ ::3L,~I +El,=. 
l.,.ll..~.l I r~: ,"E,,-I + I-', 1~: (:3 I 
L~\]~.I + + +O ,'n'~ ~+ O ~LEL,J 
~ :~;~. P_. / /0::3 r~ L) + ~L, rE 
//' ~1~ O O I"1 ) I'--I ~"~ ~" "1" LIJ h~ I X 
+t'J+ +'"~ + :~ '--' "~ a...-I 
/rh 0 /:> + l.,.I L.~ <l: <I: Oi::~ 
I.~11.,J + ~,.I i-, ..~.. ~ {/~ / + 1.~1 
noun phrase (SN). 
After the preliminary definitions, the lexicon contains a set 
of lexical entries. In a computer derivation of a sentence, lexical 
items will be selected at random from those of the appropriate cate- 
gory which match inhereht, features already in the tree and have con- 
textual features, satisfied by the tree. 
Figures ~ and 5 are presentations of the transformational 
component. In the computer version a transformation consists of 
three parts, identification, structural description (SD), and 
structural chan~e (SC). The identification contains the number and 
name of the transformation, plus information to be used in deter- 
mining when it is to be invoked. This includes a group number 
repetition parameters, keywords, etc. The structural description is 
similar to the linguistic form, but allows also subanalysis to any 
depth. Representation of the structural change by a sequence of 
elementary operations removes any possible ambiguity from the state- 
ment. In addition to adjunctions and substitutions, there are also 
elementary operations which alter complex symbols. \+PASSIF \ MERGEF 
adds a new feature specification to the complex symbol of term 4. 
\*FEM *PERS\ MOVEF 4 7 will change those two features of term 7 so 
that they are the same as for term 4. 
It may be noted that the transformation LOWESTS and the control 
program of Figure 5 have no correspondents in Figure 4. They are 
needed because the program requires that the traffic rules be given 
explicitly as part of a grammar. LOWESTS selects the lowest sentence 
\[~\] ~ST-~J 
# (P~) SN V SN (SN) p~ A # 
i 2 3 4 5 6 7 8 9 => 
l 2 8 4 5 6 7 3 9 
\[Tg\] ANTEP-OBL. OBL 
# .(P~) A v SN (sN) p~+SN # 
1 2 3 4 5 6 7 8 =~ 
i 2 5 4 <+passif ~ 6 7 8 
\[TI3\] AC-PRED 
# (PRE) \[(DET) 
l 2 3 
li fem ~ (P)\]sN pers| 
2 pers~ 
p1~ IN 
4 5 
i 2 3 4 5 
OBL 
(cop) 
COND: 7 $ ~fem 
pers 
~2 pers 
~plur 
\[T33? ELLIPSE ## 
# X 
1 2 
2 
# 
3 
Figure 4 
Transformations, from \[I\] 
OBL 
0BL 
AD 
7 
# 
9=> 
7~ ~fem 
~pers 
,~2 pers 
~plur 
9 
\[ ~l\] 
\[~52\] 
M-PASS. 
x <+~ssif> 
1 2 
l+est 2(+~ 
cow: 2~+~ 
TR- TRAITS- PASS 
X est 
1 2 
i 2 • ~progr 
( d futur 
( ~preterit 
( ~per s 
<42 pers 
<~fem 
(gplur 
(~inf 
CON\]), 2~4prog 
~futur 
~preterit 
Y 
V 
3 => 
le+ 3 
~+passif ~ 
~progr 
futur 
~preterit 
~pers 
2 pers 
a inf 
afem 
4plur 
V 
0BL 
0BL 
v 
4 => 
4 
Figure 4 (Continued) 
~0 
Figure 5 
Transformations 
v 
<~0 
oo L~ 
,<~ 
<~ 
~ 2 
: I--.. ,,~ ~ ~ 
0 "-' .-~ + "U: ~ ~" * 
V r.~ ~ .--I '-~ "~ r'-- (,") 
/ 
l"1 
0 
o C~, L~I / l.J.I 
,~, I~.. C~ C'< / ~..~ L.z. ,m-- 
(~ C.~ L~ ~....4 r~ 
(~'~ C.) <~ ,,::~ / ""t 
.--.1 r< L,...I >..,m ....I IT~ I--, C:~ 
~ ¢,'~ ~0,,1 I-.-, C%I ~J I:1... 
,m C< L,..I r,,~/ 
0 °~ 
0 
~Jt--' ~Z 
O0 
I 
LLI 
I.'-4 
.~ 
o~ 
A 
V 
(~ 
(0 
O 
...1 
Figure 6 
Base Tree, from \[1\] 
" 
V 
I I I I 
vvvvV d} 
i I "'~J 
I" ' ~ 
OJ 0.. ~ ¢-t ~ 
&'} + I I 1 I E', 
V V V V V 
t ,I 
~J 
.Q 
I ' 
r~ 
L~ 
p4 
,Q 
@ 
E-~ 
i2" 
which contains boundaries. The control program specifies that the 
cyclic transformations are to be carried out for this lowest 
sentence. After a cycle the boundaries are erased and the next 
highest sentence becomes lowest. The postcyclic transformations will 
then be carried out. 
A particular tree, created as an example in Ill, is presented 
in Figures 6 and 7. Figure 7 contains two alternative versions, a 
fixed-field format and a bracketted free-field form. Either of these 
is acceptable to the programs. The sentence at the top of the figure 
is merely a title; it will not be processed by the program. The 
lexical items "Trudeau", "deGaulle", and "berne" have not been in- 
cluded, although they could have been. If these items had been 
entered in the tree, the lexical insertion process would merely have 
added the appropriate complex symbols for them. 
Figure 8 gives the derivation as presented in Ill. Figure 9 is 
the final part of the listing of the computer output. 
The use of the ~ro6rams 
The system was designed to be used by a linguist who is in the 
process of writing a transformational grammar. As lexical items or 
transformations are added they can be tested in the context of all 
previous rules and their effect can be examined. 
The easiest errors to detect and repair in a grammar are 
syntactic errors. As a grammar is read in by the program a check is 
made for formal correctness. For each error a comment is produced 
which attempts to explain what is wrong. The program then continues 
~3 
Figure 7 
Alternative forms of Base Tree 
~ Z 
,..I 
lxJ 
t~ 
A 
^ 
A 
V 
A 
V 
U'/ 
V 
A 
V 
(fl 
V 
'-I 
r~ 
A 
V 
V 
14 
Figure 8 
Derivation, from \[ l\] 
e. 
A 
z ~.. 
.-~ ,-4 
,-4 ~ r4 
0 P. Z 
. ~/~ l-i l @ 
0 ~ E~ ..~ ~ 
~n ~ 0 Z 
> 
~ ~1~1~ ~ ~ 
+.~ I ! I I.~" v v vv 
! ^ I "z 
U I I I I I-i ,,~ v v vv~ 
! 
Z 
> 
+ 
Z 
0 ,-q 
~5 
0 
E 
0 
r~ 
AAAAAAAA 
Figure 9 
Computer Derivation 
I,d 
b4 
E 
b.I 
O! 
ILl 
(/1 
,_,1 {f) ,,~ :~ 
Z 
N 
/ 
,-I 
A ! ^ 
~ o 
V ! 
V ~* 
, .~ 
23 I 
.-.I E ~ 
* E 
L,..I 
\[tl 
Q C~ 
I O 
/{;1 / 
/ 
+ 
r,, 
t~ 
4- 
I-,, 
"l 
O ^ 
I,.I 
::~ rd'} 
I.~ V ! ! 
:3 -m 
.-I ._J 
! $ 
E E b,4 
* 4- 
I.s.I 
r~ 0.. 
O O 
! I 
! I 
+ ~ 
/ / 
2:> Z 
E ,-.., 
.-J 
I.s.I 
c~ 
r~ 
bJ 
\[.0 
L:.,I 
f~ 
.J <~ 
z 
O O 'O O 
A'AAAAAA~AAAAAAAAAA 
16 
to read in the rest of the gra~nar, recovering as best it can from 
the error. In most cases a single error will cause a small part of 
the grammar to be read badly, but the rest of the grammar will be 
read in and used in whatever tests were requested. An effort was 
made to make the error com~aents as clear and explicit as possible, 
and to make the program continue despite input errors. 
Deeper errors arise when a grammar is syntactically correct, 
but does not correctly describe the language of which it purports to 
be a grammar. ~lese errors of intent cannot be detected directly by 
the program, since it has no standard of comparison. The program 
attempts to provide enough feedback to the linguist so that he will 
be able to detect and investigate the errors. 
The information produced by the program consists of derivations 
which may be partially controlled by the user. Since random deriva- 
tions have been found to be of r@latively little interest, the system 
allows the user to control the sentences to be generated so that 
they are relevant to his current problem. (The device used for this 
purpose has been described in \[g\].) It is only in the sense of 
providing feedback to the user that the system can be called a 
"grammar tester"; it does not directly seek out errors in a gran~nar, 
nor does it evaluate the grammar. 
For a standard run of the system the inputs are a grammar, a 
t 
SMAIN card, and some trees. The grammar consists of one or more of 
phrase structure, lexicon, and transformations. The SMAIN card is 
a specification of the type of run to be made. The system must be 
i7 
told (i) what type of input trees to expect: 
TRIN, for fixed-field tree 
FTRIN, ffor free-field bracketted tree 
(2) whether to generate a tree around a skeletal input or whether it 
is only necessary to insert lexical items: 
GEN, to generate a tree and insert lexical items 
LEX, to insert lexical items 
and (3) whether or not transformations are to be applied: 
TRAN, if transformations are to be invoked. 
The general form of the SMAIN card can be represented as 
SMAIN I TRIFTR~N~ ((n)I~l)(TRAN) . 
The integer n specifies the number of time each input tree is to be 
used. 
An an example, 
$MAIN TRIN GEN TRAN . 
specifies a run in which a skeletal tree is read, a full tree is 
generated including lexical items, and the transformations are 
applied. 
The specification 
$~u~ ~I~ 5 u~x T~. 
might be used in testing a lexicon and transformations against a 
fixed base tree. The tree will be read and five cases of lexical 
/ 
~8 
insertion plus transformation will be carried out. 
SMA~N ~IN 4 nEX . 
would do four examples of lexical insertion for each input. 
After the process is completed for one input, another input is 
read and the cycle repeats. A run terminates when there are no more 
inputs. 
Computer experiments in transformational ~rammar 
The system has been in use since February 1968, although not 
fully complete at that time. The first experiments were carried out 
by the designers of the system, using granrnars based on material in 
the linguistic literature. This was done to provide test material 
for the programs, but, more importantly, to help ensure that the 
notational conventions would be adequate. A fragment of grammar 
from Chomsky's Aspects was used to test ideas and programs for 
lexical insertion. The II~ Core Grammar of Rosenbaum and Lochak 
\[6\] was used in developing and testing the transformational component. 
Both of these projects led to valuable teaching materials, as we 
shall discuss later. 
Aspects and Core provided us with separate examples of lexicon 
and transformations. There was at first no single source which con- 
tained both. A relatively formal grammar was needed, even though a 
final translation into the notation of the system would still of 
course be necessary. Elizabeth Closs Traugott's Dee~0 and surface 
structure in Alfredian Prose \[ 7 \] appeared at about that time and 
was the first grammar which was formalized in the notation after the 
19 
fact. Considerable effort had gone into designing the notation; we 
were anxious to see if it would now seem natural for a grammar which 
was new to us. Alfred was thus the first real test for the system. 
As it turned out there were a few difficulties which arose because the 
notation had not been explained clearly enough, but the results of the 
run were also revealing about the grsm~nar. 
One general effect which was noticed in these first few cases 
had continued to be striking: the need for complete precision in 
the statement of a grammar forces the linguist to consider problems 
which are important, but of which he would otherwise be unaware. 
Also during the spring of 1969 Barbara Hall Partee made two 
sets of runs with preliminary versions of a grammar of English being 
developed by the U.C.L.A. Air Force English Syntax Project. This 
grammar presented another kind of challenge to the system, because 
it was not based directly on the Aspects model, but incorporated some 
recent ideas of Fillmore. As before, these runs assisted in cleaning 
up the programs but were also of interest to the linguist. The major 
advantages from the linguistic point of view seem to have been, first, 
that the notational system of the computer model provided a framework 
in which grammars could be stated, and second, that the computer runs 
made it easier todetect certain errors in the grammars. In the main, 
these errors were not particularly subtle, and could have been caught 
by working over the grammar carefully. 
The program was also used by L. Klevansky~ who wrote a grammar 
of Swahili for the dual purposes of testing the programs and learning 
the language. 
20 
These early experiments are described in a report \[5\] which 
gives the gran~nars as well as a detailed discussion of the results 
of the computer runs. 
The form of the French grammar used in the extended example 
above is based on the form of the Core grammar; it was therefore 
easily translated into the notation of the system. Shortly after 
the grsmmnar was received, a large part of it was running on the 
computer. Minor errors in the grammar have been found and corrected; 
it will now be available to students as another example of a trans- 
formational grammar. 
The next experiment planned using the system is a project 
proposed by Susumu Nagara and Donald Smith at the University of 
Michig~, who plan to use the system to aid in writing a grammar of 
Japanese. 
Modifications to grammars based on computer runs 
In almost all cases the gran~nars used with the system have 
been sufficiently complete for at least informal distribution. The 
programs were really designed to make it easier to write grammars, 
not to test completed grammars. Nonetheless, on the basis of computer 
runs, certain types of changes have been found to be needed in the 
grammars. The cotangents which follow are based on all the grammars; 
they do not all apply to any one of them. 
i 
Trivial corrections 
The most co~on errors are typographical errors in transcription 
of the grammar. These are not errors in the grammar itself; having 
2i 
to deal with them is one of the prices of using the computer. In 
general, these can be caught with relative ease. 
More than one grammar has had simple errors with respect to 
repetition of a transformation. Number agreement transformations 
are written so that they produce CAT S S S ... where CAT S is wanted. 
(The grammar as written calls for an infinite sequence of S's to be 
added. The program, more cautious, adds ten S's, then complains and 
goes on to the next transformation. ) 
Transformations are often stated so that they fail to apply in 
all cases where it is intended they apply. For example, the 
structural description of PASSIVE as 
SD # (PRE) 3NP AUX 5V (PREP) 7NP % PREP 10P % # , 
WHERE 3EQ7. 
fails to take into account some additional parts of the VP. The 
correction to 
SD # (PRE) ~NP AUX (HAVE EN)(BE ING) 5V (PREP) 7NP 
PREP lOP ~ #, WHERE 3 EQ 7- 
will allow PASSIVE to work in the additional cases. Similarly, a 
NOMINAL-AGREemeNT transformation which marks subjects as +NOMIN must 
apply not only to pronouns which precede verbs but also to those which 
precede copulas. Thus the structural description 
SD # ~ 3(~ON, REL) V ~ # . 
must be replaced by 
SD # ~ 3(PRON, REL) (V, COP) % # . 
22 
Interrelatedness of transformations 
A slightly more interesting set of problems found in the 
computer runs are those which arise through the interrelatedness of 
two or more transformations. For example, in one of the grsmm~ars 
there ~ere both WH-questions and TAG-questions. It was found that 
the TAG transformations was (optionally) applicable to any question, 
so that for example 
TOM HAS PREFER EN WHAT GIRL HAS TOM NOT 
was produced. This error was easily repaired once it was detected. 
On the other hand, a similar problem which was not easily 
fixed arose with another transformation which was marked optional. 
Testing showed that for certain base trees the ~esult was bad if the 
tr~usformation did not apply; however3 when the transformation was 
l 
temporarily changed to obligatory, the grammsx then failed to produce 
some intended sentences. The proper correction to the grammar would 
have required specification of the contexts in which the transforma- 
tion was obligatory. 
Incompleteness of grammars 
Formal gram~nars so far have each attempted to describe some 
subset of a language. In computer testing many problems outside 
the scope of the grammar are evident. If, for example, a grammar 
does not treat prepositions seriously, then once this becomes apparent, 
i 
the computer runs need to be designed to avoid prepositions. 
Dee~ structure ~roblems 
Two of the grammars which have been studied suffer problems 
Z3 
with the WH-morpheme when it occurs in non-sentences and not as a 
relative marker. Thus, for example, sentences such as 
WHAT BLAME MAY NT BE BE ING 
and 
WHICH THING MUST HAVE BE EN APPROVE ING OF 
WHAT TABLE 
are in fact even worse than they appear, because they are not 
questions. Although this problem has no simple solution in the 
current framework, the inputs to the program can be controlled to 
avoid generating sentences of this form. 
Inadequacies in the linguistic model 
An interesting change to the system was suggested by the 
attempt to formalize the Core grammar. In both the WH-attraction 
and the Question-transformations the structural description contains 
a two-part choice between a PREP NP pair and simply an NP. This is 
of the form: 
% (PREP NP, ~P) 
where ~ is a variable. Any structure which satls~ies 
the first part of the choice will also satisfy the second, and any 
analysis algorithm must have some order of search which will either 
always select PREP NP or always select NP only. But the intent is 
that there should be a genuine choice, so that the grammar produces 
both 
ABOUT WHAT DID JOHN SPEAK 
and 
24 
WHAT DID JOHN SPEAK ABOUT 
The solution which was found for the problem was to add an additional 
value (AAC) for the repetition parameter for a transformation. 
If a transformation is marked AAC, all possible analyses will 
be found, but only one of them, selected at random, will be used as 
the basis for structural change. This seined the appropriate way to 
solve the problem for the Core grammar, and it turned out also to 
solve a slightly different repetition problem in the grammar of A1- 
fredian prose. Notice that this is really an observation about the 
form of grammars, rather than about a particular grammar. Yet it 
arose by consideration of particular examples. 
Surface structure 
The surface Structure associated with a sentence derivation is 
much easier to study if it can be produced automatically. In several 
cases it has been apparent from the information provided by the computer 
runs that revisions in the grammar were needed if the surface structure 
is to be at all reasonable. This is a case where the computer runs are 
certainly not necessary, but where they reduce the tediousness of 
studying the problem. 
In stmmmary, it seems to me that main value in computer testing 
of a completed grsm~nar is that the need for a precise statement 
brings to the consideration of the linguist problems which are other- 
l 
wise below the surface. These problems may be in the grammar itself 
or they may be in the linguistic model itself. For a grammar in 
process of being written the greatest advantage is in allowing rules 
25 
to be checked as they are added, and in bringing out the interaction 
between rules. 
Instructional use of the s~stem 
The system has now been used by Sziliard Szabo in teaching 
~eneral linguistics at the University of San Francisco, by Michael 
O'Malley in a course in natural language structure at the University 
of Michigan, and by the author in courses in co~0utational linguistics 
at Stanford and Michigan. 
The method of use is to make available to the students a file 
of one or more grammars to be used as examples and as bases for 
modifications. The fragments from Aspects and the IEM Core grammar 
have been most useful3 although small grammar written for this purpose 
have also been used. The students are then asked to make modifications 
and additions to the grammars. 
For graduate students, a reasonable exercise for a term paper 
is to read a current journal article on transformational grammar, and 
then show how the results can be incorporated into the basic grammar, 
or show why they cannot be. The papers chosen by the students have 
generally been ones in which transformations are actually given. 
This project has been very successful as am introduction to trans- 
formational grammar for computer science students. 
Other students have chosen simply to use the computer to obtain 
fully developed examples of derivations illustrating aspects of 
grammar in which they are interested. 
These experiences have confirmed our belief that specific 
26 
examples presented by the computer, and the feedback provided when 
a student modifies a grammar, are valuable in enabling the udent 
to understand the notion of trausformational grammar. 
27 

References 

\[i\] Colmerauer, C., M. Courval, M. Poirier, and Antonio A. M. Querido. 
Grammaire- I~ Description s~ntaxi~ue d'un sous-ensemble du francais, 
Universite de Montreal, (March3 1969). 

\[2\] Chomsky, N. Aspects of the Theory of S~ntax. M.I.T. Press, 
Cambridge, Massachusetts (1965). 

\[3\] Friedman, Joyce. A computer system for transformational grammar. 
Cosnn. ACM (to appear). 

\[4\] Friedman, Joyce. Directed random generation of sentences. Comm. 
AC_~M, 12, pp. 4O-46. 

\[5\] Friedman, Joyce. (Ed.) Computer Experiments in Transformational 
Grammar, CS-108, Computer Science Dept., Stanford University, 
(August, 1968). 

\[6\] Rosenbaum, R., and Lochak, D. The I~M core grammar of English. 
In Lieberman, D. (Ed.) Specification and utilization of a 
transformational grammar. AFCRL-66-270 (1966). 

\[7\] Traugott, Elizabeth C. Deep and surface structure in Alfredian 
prose. Mimeographed. PEGS Paper #14 (August, 1967). 
