Relational-Grammar-Based Generation in the JETS Japanese-English 
Machine Translation System 
David E. Johnson 
IBM Research, T. J. Watson Research Laboratory 
P.O. Box 218 
Yorktown Heights, NY 10598 USA 
Hideo Watanabe 
IBM Research, Tokyo Research Laboratory 
5-19 Sanbaneho, Chiyoda-ku 
Tokyo 102, Japan 
Abstract 
This paper describes the design and func- 
tioning of the English generation phase in 
JETS, a limited transfer, Japanese-English 
machine translation system that is loosely 
based on the linguistic framework of relational 
grammar. To facilitate the development of 
relational-grammar-based generators, we have 
built an NL-and-application-independent gen- 
erator shell and relational grammar rule- 
writing language. The implemented generator, 
GENIE, maps abstract canonical structures, 
representing the basic predicate-argument 
structures of sentences, into well-formed 
English sentences via a two-stage plan-and-ex- 
ecute design. This modularity permits the 
independent development of a very general, 
deterministic execution grammar that is driven 
by a set of planning rules sensitive to lexical, 
syntactic and stylistic constraints. Processing 
in GENIE is category-driven, i.e., grammatical 
rules are distributed over a part-of-speech 
hierarchy and, using an inheritance mech- 
anism, are invoked only ff appropriate for the 
category being processed. 
1- Introduction 
This paper discusses relational-grammar-based gener- 
ation in the context of JETS, a Japanese-English 
machine translation (MT) system that is being devel- 
oped at the IBM Research Tokyo Research Laboratory. 
To put our work in perspective, we first explain the 
motivation for basing JETS on relational grammar (RG) 
and then sketch the processing flow in translation. With 
this background, we (i) describe and illustrate certain 
aspects of the rule-writing language, GEAR, in which 
the GENIE English generator has been written; (ii) 
comment on key aspects of the generator shell, GEN- 
SHELL, in which GENIE has been developed; and (iii) 
discuss the design and functioning of the GENIE 
English generator. 
With few exceptions such as the work being done 
at CMU (cf. KBMT-89 (1989), Nirenburg (1987), and 
Nirenburg, et. al. (1988)), in the SEMSYN project at 
the University of Stuttgart (Rosner (1986)), and the 
joint work between the ISI Penman project and the 
University of Saarbrticken (Bateman, et. al. (1989)), 
generation within the area of machine translation has 
received very little attention. Typically, MT systems 
have no independently functioning, linguistically justi- 
fied generation grammar. In the case of transfer 
systems, much of the target language grammar is typi- 
cally built into the transfer component, resulting in a 
non-modular, rigid and linguistically inadequate system. 
It is the norm in MT systems for the linguistic complexi- 
ties inherent in robust generation to be simply ignored, 
contributing to the inadequacy of MT systems. 
In contrast, we have sought to shift more of the 
processing burden from transfer onto generation, 
allowing our system to incorporate a variety of results 
coming from theoretical linguistics. GENIE is an appli- 
cation-and-language-independent generator embodying 
174 
a robust, linguistically justified RG grammar of English. 
Moreover, GENIE incorporates a syntax planner that 
applies a set of planning rules determining which rules 
in the execution grammar should be applied. As long 
recognized in work on text generators, the incorpo- 
ration of a syntax planner introduces the kind of flexi- 
bility required for robust generation. 
JETS is a so-called limited transfer system, i.e., a 
system in which structural transfer is kept to a 
minimum. The key RG notion in our work is that of 
canonical (relational) structure (CS), an abstract level of 
syntactic structure representing the basic predicate-ar- 
gument structure of clauses in terms of a universal set of 
primitive (grammatical) relations such as subject, direct 
object, indirect object, chomeur. 1 
Given the basic assumption that one is developing 
a limited transfer system, implying deep analyses of 
both the source and target languages which converge on 
structurally similar internal representations for trans- 
lation equivalents in a wide range of cases, it is critical 
to select a linguistic framework which supports the 
required analyses, enabling one to conceptualize the lin- 
guistic processing in a uniform manner. As discussed in 
Johnson (1988b), with respect to MT, RG is a logical 
choice of linguistic framework since CSs provide a 
natural syntactic bridge between languages as diverse in 
structure as Japanese and English. This is so for two 
reasons: (1) within one language, the CSs of para- 
phrases are typically the same or highly similar and (2) 
translation equivalents often have structurally similar if 
not isomorphic CSs. 
One of the key advantages of RG comes from its 
explicit representation of grammatical relations like 
subject and direct object, which are argued to be uni- 
versal. In contrast, structure-based frameworks such as 
transformational-generative grammar (TG) at best only 
implicitly represent grammatical relations such as 
subject and direct object in terms of linear precedence 
and dominance, which are language particular. If one 
considers the task of transfer, for instance, it is clear 
that representing basic clause structure in terms of 
explicitly marked, order-independent relations rather 
than in terms of language-dependent structural relations 
reduces the amount of structure changing to be done in 
the transfer component. This is especially true for lan- 
guages like Japanese and English, which differ greatly in 
superficial structural properties (not to mention the fact 
that Japanese has very free word order, which arguably 
makes it even less suited to structure-based frame- 
works). 
2- Processing Flow in JETS and GENIE 
As in all transfer systems, linguistic processing in JETS 
can be divided into three phases: analysis, which con- 
sists of lexical analysis and parsing, transfer and gener- 
ation. The output of analysis is a Japanese CS, which 
represents the basic predicate-argument structure of the 
Japanese sentence. 2 Transfer produces an English CS, 
which is often, but not always, isomorphic to the Japa- 
nese CS. The English CS is passed to the GENIE gen- 
erator, whose task is to generate a grammatically 
correct and stylistically appropriate English sentence 
given a well-formed CS. 
To illustrate, consider the following Japanese sen- 
tence and two of the possible English translations: 
1. karera wa Tookyoo e itta rashii 
they top Tokyo to went seem 
2. They seem to have gone to Tokyo. 
3. It seems that they went to Tokyo. 
In translating (1), analysis maps the input string into the 
Japanese CS shown at the left in Figure 1 on the next 
page. Transfer then maps the Japanese CS into the 
English CS shown at the right in Figure 1. 
I For theoretical background on RG, see the many articles listed in the bibliographic reference Dubinsky and Rosen (1987). 
Note that the following abbreviations are used in glosses of Japanese examples: top (topic), nm (nominalize), and pp (post- 
position). 
2 For discussion of parsing in JETS, see Maruyama, Watanabe and Ogino (1989). 
175 
Japanese CS for (1) English CS for (2) & (3) 
rashii seem 
itta go(tense, past) 
/loci 1/ loci 
karera Tookyoo they Tokyo 
(topic. wa) (pp. e) (topic. T) (prep. to) 
seem 
go(past) 
loci 
Tokyo(to) 
Tense- 
Spelling 
they 
seem 
have they 
go(pastpart) ,< 
Tokyo(to) 
Figure 1. Canonical structures for (1), (2) and (3). Note 
that "1" means "subject", "loc" means "locative". 
Given the English CS, it is up to the GENIE English 
generator to generate either (2) or (3). Based on the 
information that they in the English CS is marked as the 
topic of the sentence, GENIE will map the CS into the 
superficial (unordered) relational structure shown in 
Figure 2 via the relational rule of Subject-to-Subject 
Raising (so-called A-raising). Subsequent rules of 
Tense-Spelling and Linearization (including the spelling 
out of verbal forms and prepositions) will result in the 
string They seem to have gone to Tokyo, as shown in 
Figure 3. 
seem seem 
go(past) go(past) they 
they Tokyo(to) Tokyo(to) 
Figure 2. A-Raising Applied to the CS of (2) and (3). 
Note that "6" means "complement". 
--- Lineanzation, etc .... > 
They seem to have gone to Tokyo 
Figure 3. Rest of the Derivation of (2) 
As illustrated above, RG, like TG, is a "multi- 
stratal" theory, i.e., clauses typically have more than 
one level of syntactic analysis, and these levels/strata 
are mediated by clause-level rules. In the case of TG, 
the structures are phrase-structure trees, and transf- 
ormations map trees into trees; in the case of RG, the 
structures are edge-labelled trees (called relational struc- 
tures (RS)), where the edge labels represent primitive 
relations, and the rules map RSs into RSs. 
The use of multiple strata sets RG apart from 
functional frameworks such as FUG (Kay 1979) and 
LFG (Bresnan 1982), which also use primitive relations 
(functions), and from all other monostratal frameworks 
such as GPSG (Gazdar, et. al. 1985), whether func- 
tional or not. The manipulation of explicitly marked 
relations in unordered relational structures sets RG 
apart from TG. In our work on Japanese-English MT, 
the RG concept of multiple relational strata has proven 
to be of significant practical use -- facilitating the 
design and development of a limited transfer component 
and a robust generation component, enhancing modu- 
larity, and allowing the linguistic processing to be con- 
ceptualized in a uniform fashion. 
176 
3- The RG Rule Writing Language: GEAR 
One key aspect of our implementation of an RG gener- 
ator is the GEAR rule-writing language. GEAR permits 
a grammar developer to write computationally powerful 
RG rules in a linguistically natural manner. GEAR 
rules identify grammatical objects via path specifica- 
tions, of which there are two types: (1) node-specifier, 
consisting of a sequence of one or more relation names, 
and (2) property-specifier, consisting of a node-specifier 
followed by a property name. For instance, 1:1 indi- 
cates a node that is the subject of a node that is the 
subject of the node currently being processed (the 
focus) and 2.tense denotes the value of the property 
tense of a node that is the direct object of the focus. 
GEAR path expressions are superficially similar to the 
expressions used in unification-based frameworks such 
as FUG and PATR (Shieber, et. al. (1983)). However, 
GEAR is not unification based, rather it provides a 
number of procedural operations, including node 
deletion and node creation. 
Each rule consists of a sequence of statements, of 
which there are several types, e.g., IF-THEN-ELSE, 
CALL, ON and restructuring statements. IF-THEN- 
ELSE statements control the rule internal processing 
flow. CALL statements are used to invoke rules by 
name. An ON statement invokes a specified rule on a 
node reachable from the focus via a node-specifier. 
There are several types of restructuring state- 
ment, e.g., ASSIGN, CREATE, DELETE and COPY. 
An ASSIGN statement is used to alter the relations of a 
node identified via a node-specifier; the new relation is 
also specified by a node-specifier. The core of 
GENIE's A-raising rule, whose relational changes are 
illustrated in Figure 2 above, is (using 6 for "comple- 
ment"): 
(ASSIGN 1 6) "Assign my subject as my complement" 
(ASSIGN 6:1 1) "Assign my complement's subject as 
my subject" 
The complete rule is shown in Figure 4. 
% % Define the rule A-raising for intransitive verbs 
(DEF-RULE A-Raising OF Intransitive-verb 
% % If the A-raising rule switch is turned on 
(IF (A-raise is 'yes) 
% % then assign my subject as my complement 
THEN (ASSIGN 1 6) 
% % and assign my complement's subject as my subject 
(ASSIGN 6:1 1) 
% % and on my complement call the rule 
% % which makes infinitives 
(ON 6 (CALL Make-lnf'mitive)))) 
Figure 4. GENIE's A-Raising rule 
Creation, copying and deletion of nodes are also 
specifiable but space limitations preclude discussion. 
4- The GENSHELL generator shell 
Building on our experience with an earlier prototype 
developed by Schindler (1988), we have developed an 
NL-independent generator shell, GENSHELL, to facili- 
tate the development of RG generators. For any given 
generator, grammar developers need only specify the 
designated grammatical relations, parts of speech, a 
part-of-speech hierarchy, dictionaries and grammars. 
GENSHELL takes this information and constructs a 
runtime generator. 
One of the distinctive aspects of GENSHELL, 
due to Sehindler (1988), is the concept of category- 
driven processing. In category-driven processing, parts 
of speech are represented as categories in a category 
hierarchy (POSH) and nodes in RSs are represented as 
objects which are instances of categories and thus can 
inherit properties via the POSH, Among the inheritable 
properties are grammar rules. For instance, the rules 
for Passive and Subject-to-Object Raising (so-called 
B-Raising; discussed later) would be associated with the 
class Transitive Verb, A-raising would be associated 
with the class Intransitive Verb, and Subject-Verb 
Agreement would be associated with the superordinate 
class Verb. 
In our implementation, all rules are defined with 
respect to named rule bundles, and rule bundles are 
associated either with categories in the POSH, the 
general/default eases, or with lexical entries, the special 
cases. Rule definitions have the form: 
I_77 
(DEF-RULE rulename OF rule-bundle-name 
(rule-body)). 
(As shown in Figure 4 above, a default rule bundle 
associated with a POS class is given the same name as 
that class.) When a node N associated with category C 
and lexical entry L is being processed, the rule search 
routine, given a rule named R -- the'latter comes from 
so-called agenda rules which are also associated with C 
D uses inheritance to first search for R among any rule 
bundles named in L, then searches for R among C's 
rules, then C's parent's rules and so on up to the top of 
the hierarchy until either some rule named R is found or 
the top category is reached and the process fails. In 
short, in category-driven processing, the grammar 
invoked on N is constructed as appropriate at proc- 
essing time on the basis of lexically activated rules and 
the rules accessible to N's category using the POSH and 
inheritance. 
One example is the ordering of adjectives and 
nouns. The class Noun is associated with a 
general/default lineanzation rule which orders adjec- 
tives before nouns, generating phrases like tall woman. 
Nouns like someone, anyone, etc. are associated with a 
lexically triggered lineafization rule which places the 
adjective after the head noun. These two rules are both 
named Linearize. Thus, if the focus is someone and it is 
modified by tall, the search routine, looking for 
Linearize, will first find the special rule, correctly gener- 
ating someone tall. 
A category-driven system has two advantages 
over more conventional rule systems: (i) it provides a 
natural mechanism for dealing with special cases trig- 
gered by lexical items, while providing a fail-soft mech- 
anism in the form of the general rules inherited from the 
POSH and (ii) only rules that in principle could be rele- 
vant to processing a given node in an RS will be tested 
for application. That is, the POSH provides a linguis- 
tically motivated means for organizing a large grammar 
into subgrammars. 3 
5- GENIE: the English generator 
Generating from CSs requires a robust generation 
grammar of the target language, as well as a decision- 
making component that decides which surface form is to 
be generated. The generation grammar employed in 
GENIE is a (deterministic) relational grammar having a 
substantial number of clause-level rules which alter 
grammatical relations, e.g., Passive, A-raising and 
B-raising, as well as minor rules such as Tense-Spelling 
and Linearization (the latter of which does not alter 
grammatical relations). 
As illustrated in Figure 1 above, CSs typically do 
not correspond directly to grammatical sentences. 
Further, any given CS typically constitutes the basis for 
the generation of a number of superficial forms, e.g., 
(2) and (3) above. This control problem has been 
addressed by splitting generation into two phases: a 
syntax planning phase and an execution phase. The 
function of GENIE's planner is quite different from that 
of other generators. Typically, generator planners 
decide "what to say", constructing some sort of internal 
representation that is then processed by a realization 
component. Typical planners will be concerned with 
chunking into sentences, topic selection and word 
choice (see, e.g., Appelt(1985), Danlos (1984), 
Hovy(1985), Kukich (1983), McKeown (1985), McDo- 
nald (1984)), and Mann (1983)). 
In the case of JETS, however, since we are in the 
domain of transfer-based MT, all of these "high level" 
considerations are decided by the analysis and transfer 
components. In GENIE's case, the planner must, on 
the basis of a given CS, deal with a myriad of low-level 
syntactic conditions and their interactions (most of 
which have not been discussed or even recognized in 
the generation literature). Internal to GENIE, this 
means deciding which of the rules in the deterministic 
execution grammar should be applied. For instance, 
CSs with seem have a disjunctive grammatical condi- 
tion: they must either be raised, yielding the pattern NP 
seem to VP (as in (2) above) , or extraposed, yielding 
the pattern It seems that S (as in (3) above). Failure to 
apply either A-raising or so-called It-Extraposition 
3 Earlier work using a lexical hierarchy and inheritance in natural language processing includes Wilensky (1981), Jacobs 
(1985) and Zernik and Dyer (1987). These works make heavy use of phrasal patterns (so-called pattern-concept pairs) and 
so the conception of grammar and lexicon and hence the notion of what is inherited in these works differ greatly from ours, 
which is part of the generative-linguistic tradition. 
178 
would result in the ungrammatical pattern *That S 
seems (in the case of Figure 1 above: *That they went to 
Tokyo seems). The decision to apply A-raising in the 
above example is stylistic ("make the topic the main 
clause subject, if possible"), but the disjunctive require- 
ment ("apply either A-raising or It-Extraposition") is 
grammatical. Having no control over "what to say", 
GENIE's planner is conceptually part of the realization 
phase and not part of the typical "planning phase". 
GENIE's planner communicates which rules 
should be applied to the execution grammar via a set of 
so-called rule switches, which are simply binary-valued 
properties whose property names are the names of exe- 
cution rules, e.g., (A-raise . Yes), (Passive . No). As 
shown in Figure 4 above, IF statements are often used 
to test for a rule-switch value, which value is either set 
by a planning rule or comes from a lexical entry. Rule 
switches are a generalization of the earlier concept of 
transformational rule features (cf. Lakoff 1970); the 
generalization is that rule switches can be dynamically set 
by planning rules, based on lexicul, syntactic, semantic and 
stylistic considerations (see Johnson 1988a for more 
examples and further discussion).'* 
For example, in (1) above, based on the informa- 
tion that they is the topic (this information comes from 
transfer), a syntax planning rule which is partly respon- 
sible for making topics surface subjects sets the switch 
(A-raise . Yes), turning on A-raising, and the switch 
(It-Extra. No), turning off It- extraposition, resulting in 
(2) rather than (3). GENIE's architecture is shown in 
Figure 5. 
Planning rules insure that a multitude of lexico- 
syntactic and stylistic conditions are met, e.g., that 
clauses with modals do not undergo A-raising, pre- 
venting the generation of, e.g., *They seem to can swim; 
that clauses with verbs like force have passivized subor- 
dinate clauses where required to meet coreferential 
deletion conditions (cf. She forced him to be examined 
by the doctor, *She forced him (for) the doctor to 
examine him); and that verbs like teach undergo dative 
alternation if there is no specified direct object, gener- 
ating He taught her rather than *He taught to her (cf. 
sing, which has the opposite condition - He sang to her 
but *He sang her). 
It is also the responsibility of the planner to make 
sure island constraints are not violated. For instance, if 
a wh-nominal is in a sentential subject, then planning 
rules turn on execution rules such as A-raising resulting 
in sentences like Who is likely to win (via A-Raising) 
rather than *Who is to win likely? or the stylistically 
marginal ?Who is it likely (that) will win?. This heuristic 
planning rule also insures that in the case of so-called 
Tough-Movement sentences, GENIE will generate sen- 
tences like Who is easy to please?, (via Tough-Move- 
ment) rather than either *Who is to #ease easy? or 
?Who is it easy to please?. 
Engli sh CS (Transfer Output) 
English CS 
(dictionary information added) 
\[ Syntax Planner I~ ~w~t~che 
English CS (rule set) 
RG Execution Grammar 
- Precycle 
- Cycle 
- Post-cycle 
- kinearization 
English Sentence 
Figure 5. GENIE Components. Note that the POSH 
contains the agenda rules and the default planning and 
execution rules organized by POS. 
4 After completing this work, we discovered that Bates and Ingria (1981) also used a mechanism similar to our "rule 
switches" to control generation within a TG framework. Their transformational constraints, however, were set by a human 
who wished to test what a given set of constraints would produce. That is, their system had no syntax planner which would 
evaluate a given base structure via a set of planning rules and set constraints insuring the generation of only grammatical 
sentences. 
179 
Execution rules are turned on (or off) either by 
syntax planning rules or by lexical entries. To illustrate 
the use of lexical rule-switches, consider the following 
example from JETS involving verbs of prevention: 
4. kanojo wa kare ga iku no o habanda 
she top he pp go nm pp prevent 
5. She prevented him from going. 
On the Japanese side, the postposition ga marks the 
subject of the embedded clause kate ga iku, which has 
been nominalized with the dummy noun no, which 
carries the direct object marker o. Following the argu- 
ments given in Postal (1974), we assume that prevent is 
a so-called B-raising trigger (B-raising is the controver- 
sial rule which relates sentences such as He believes that 
she knows (not raised) and He believes her to know, in 
which her is raised up as direct object of believe). The 
CS for (5) is as shown to the fight in Figure 6 and the 
CS of the Japanese sentence (4) is shown to the left: 5 
Japanese CS for (4) English CS for (5) 
habanda 
kanojo iku 
(topic. wi/ 
kare 
prevent 
TRANSFER /2~ 
she go 
(topic. T) 1/ 
7 
he 
Figure 6. Canonical Structures for (4) and (5) 
GENIE's rule of B-raising, given in Figure 7, maps the 
English CS into a superficial RS, as shown in Figure 8. 
As shown in Figure 6, the English and the Japanese 
CSs are isomorphic, i.e., there are no structural changes 
in transfer. 
To produce (5) from the English CS in Figure 6, 
as illustrated in Figure 8, merely requires the dictionary 
entry depicted in Figure 9. 
% % Define the rule B-raising for transitive verbs 
(DEF-RULE B-Raising OF Transitive-Verb 
%% If the B-raising rule switch is "yes" 
(IF (B-raise is 'yes) 
% % then make my direct object my complement 
THEN (ASSIGN 2 6) 
% % and make my complement's subject 
% % my direct object 
(ASSIGN 6:1 2) 
% % and on my complement call the rule 
% % that makes infinitives 
(ON 6 (CALL Make-lnf'mitive)))) 
Figure 7. GENIE's B-Raising Rule 
prevent prevent 
- .... >Yl 
s e/o :L go 1 
(prep. from) 
he (ccomp. ing) 
= Other Rules = > She prevented him from going 
Figure 8. Example of B-Raising Application 
:lexical-form. prevent 
:category. transitive-verb 
:rep-lexical-form. nil 
:rep-category. nil 
:properties. (B-Raise. Yes) (cprep. from)(cvform, ing) 
:additional-rule-sets. nil 
Figure 9. Lexical entry for "prevent" 
This lexical entry states that prevent is a transitive verb, 
hence has access to the rules defined for transitive verbs 
s Postal's English-internal arguments were based on the fact that the direct object of prevent could be existential there, 
weather it and idiom chunks (cf. She prevented there from being a riot/it from raining/the cat from being let out of the bag). 
180 
in the POSH, e.g., Passive and B-raising (and the rules 
of superordinate classes), and that among its properties 
are the rule switch setting (B-Raise . Yes), which trig- 
gers Subject-to-Object raising, the feature (ccomp . 
from), which determines that the complement clause 
(fragment) will be flagged with from via a general rule, 
and the feature (cvform . ing), which Make-Infinitive 
will use when called by B-Raising to determine the verb 
form going in the example. Prevent has no 
rep(lacement)-lexical-form, which is used, e.g., to map a 
single input form such as look-up into a verb look and a 
particle up, or more generally to map senses into lexical 
strings. "Rep-cat", also nil here, can be used to map 
one category system into another (not used in GENIE). 
"Additional-rule-sets", also nil, is the repository for the 
names of any rule bundles associated with a lexical 
entry (e.g., easy, hard, etc. would have the additional- 
rule-set name tough-movement, which contains the 
Tough Movement rule and the planning rule that turns 
Tough Movement on). 
As depicted in Figure 5 above, the execution 
component consists of three relation-changing phases, 
called "pre-cycle", "cycle" and "post-cycle", in which 
execution rules are applied bottom-to-top, followed by 
a top-down linearization phase, which builds an output 
string that is then sent to the morphological component 
(not shown). Each phase has its own set of agenda 
rules, whose functions are to either call grammatical 
rules or shift control, i.e., agenda rules are a sequence of 
CALL statements. Agenda rules, like grammatical rules, 
are defined for classes, so that, e.g., the cyclic agendas 
for adjectives, nouns and verbs are different. For 
instance, part of the agenda for the cyclic phase of tran- 
sitive verbs is: ... (Call B-raising) (Call Dative) (Call 
Passive) .... but none of these rules are relevant to 
adjectives, nouns or intransitive verbs. It should be 
noted that rules called by a particular agenda might be 
accessed via inheritance. E.g., Reflexivization is called 
in the cyclic agenda for transitive verbs, but it is associ- 
ated with the class Predicate so that it is available to 
adjectives in cases like He is proud of himself (it is 
assumed that Reflexivization is executed on the proud 
clause before A-Raising applies on be). 
The grammar implemented in GENIE to date 
includes many of the important rules for English clause 
structure, including Yes/No questions, Wh-questions, 
relative clauses, subordinate clauses of various types, 
verb-particle combinations, raisings of various sorts, 
passives, and extrapositions. 
6- Concluding Remarks 
We have developed an application-and-NL-independent 
generator shell, GENSHELL, including a flexible dic- 
tionary system and a high-level rule-writing system, 
GEAR, to facilitate the development of category-driven 
RG generators. GENSHELL/GEAR provides a pow- 
erful computational framework for the development of 
RG-based natural-language-processing components. 
We have also implemented GENIE, a robust English 
generator, within GENSHELL/GEAR. Besides the 
novel use of RG and category-driven processing, 
GENIE is notable for its two-stage plan-and-execute 
design. 
JETS and GENIE are currently being tested on 
sentences from Asahi newspaper editorials on economic 
matters, a challenging task since editorial sentences can 
be very long, with essentially unrestricted vocabulary. 
Nevertheless, we have found the initial tests of the gen- 
erator encouraging, supporting our view that besides its 
intrinsic theoretical interest, RG has practical value in 
natural language processing. 

References 
Appelt, D. E. 1985. Planning English Sentences. ACL 
Series: Studies in Natural Language Processing. 
Cambridge UP, Cambridge. 
Bateman, J., R. Kasper, J. Schtttz and E. Steiner. 1989. 
"Interfacing an English Text Generator with a 
German MT Analysis," submitted to the European 
ACL, Manchester, 1989. 
Bates, M. and R. Ingria. 1981. "Controlled Transfor- 
mational Sentence Generation," Proceedings of the 
19th Annual Meeting of the Association for Compu- 
tational lJnLqlistics, Stanford, CA. 
Bresnan, J. (ed.) 1982. The Mental Representation of 
Grammatical Relations. MIT Press, Cambridge, 
Mass. 
Danlos, L. 1984. "Conceptual and Linguistic Decisions 
in Generation," Proceedings of COLING-84, Stan- 
ford, pp. 501-504. 
Dubinksy, S. and C. Rosen (eds) 1987. "A Bibli- 
ography on Relational Grammar Through 1987 with 
Selected Titles on Lexieal-Functional Grammar," 
distributed by Indiana University Linguistics Club, 
Bloomington, Indiana. 
Gazdar, G., E. Klein, G. Pullum and I. Sag. 1985. Gen- 
eralized Phrase Structure Grammar. Harvard Univer- 
sity Press, Cambridge, Mass. 
Hovy, E. 1985. "Integrating Text Planning and Pro- 
duction in Generation," Proceedings of IJCAI-85, 
Los Angeles, CA. 
Jacobs, P. S. 1985. A Knowledge-Based Approach to 
Language Production. PhD dissertation, UC 
Berkeley, Computer Science Division, UCB/CSD 
86/254, Berkeley, CA. 
Johnson, D.E. 1988a. "On the Linguistic Design of 
Post-Analysis in the JETS Japanese-English 
Machine Translation System", Proceedings of the 
International Conference on Fifth Generation Com- 
puter Systems 1988, Tokyo. 
Johnson, D. E. 1988b. "A Relational Grammar 
Approach to Machine Translation," Proceedings of 
the Information Processing Society of Japan, Natural 
Language Processing, Vol. 88.61. 
Kay, M. 1979. "Functional Grammar," Proceedings 5th 
Annual Meeting of the Berkeley Linguistics Society, 
Berkeley, CA, pp. 142-158. 
KBMT-89. 1989. KBMT-89 Project Report, Center 
for Machine Translation, Carnegie Mellon Univer- 
sity. 
Kukich, K. 1983. Knowledge-Based Report Generation: 
A Knowledge Engineering Approach to Natural Lan- 
guage Report Generation. Phi) dissertation, Infor- 
mation Science Department, University of 
Pittsburgh. 
Lakoff, G. 1970. Irregularity in Syntax. Holt, Rinehart, 
Winston, New York. 
Mann, W. 1983. "An Overview of the Penman Text 
Generation System," Proceedings of the National 
Conference on Artificial Intelligence, pp. 261-265. 
Maruyama, H., H. Watanabe, and S. Ogino. 1989. "An 
Interactive Japanese Parser for Machine Trans- 
lation," Proceedings of COLING90, Helsinki, to 
appear. 
McDonald, D. 1984. "Description Direct Control: Its 
Implication for Natural Language Generation," in 
N. J. Cercone (ed.) Computational Linguistics, Per- 
gamon Press, Oxford, pp. 403-424. 
McKeown, K. 1985. Text Generation. Cambridge Uni- 
versity Press, Cambridge. 
Nirenburg, S. 1987. "A Distributed Generation System 
for Machine Translation," Technical Report, Center 
for Machine Translation, Carnegie Mellon Univer- 
sity. 
Nirenburg, S., R. McCardell, E. Nyberg, P. Werner, S. 
Huffman, E. Kenschaft, and I. Nirenburg. 1988. 
"Diogenes-88," Technical Report, Center for 
Machine Translation, Carnegie Mellon University. 
Postal, P. M. 1974. On Raising. MIT Press, Cambridge. 
ROsner, D. 1986. "When Mariko Talks to Siegfried - 
Experiences from a Japanese/German Machine 
Translation Project," Proceedings of COLING-86, 
Bonn. 
Schindler, Peter A. 1988. "General: An Object-Or- 
iented System Shell for Relational Grammar-Based 
Natural Language Processing", master's thesis, 
Department of Electrical Engineering and Computer 
Science, MIT. 
Shieber, S. M., H. Uszkoreit, F.C.N. Pereira, J.J. 
Robinson and M. Tyson. 1983. "The Formalism 
and Implementation of PATR-II," In Research on 
Interactive Acquisition and Use of Knowledge, AI 
Center, SRI International, Menlo Park, CA. 
Wilensky, R. 1981. "A Knowledge-Based Approach to 
Natural Language Processing: A Progress Report," 
Proceedings Seventh International Joint Conference 
on Artificial Intelligence, Vancouver. 
Zernik, U. and M. G. Dyer. 1987. "The Self-Ex- 
tending Phrasal Lexicon," Computational Linguistics, 
vol. 13, No. 3-4, pp. 308-325. 
