REVERSIBILITY AND MODULARITY IN 
NATURAL LANGUAGE GENERATION 
Gfinter Neumann 
Lehrstuhl ffir Computerlinguistik 
SFB 314, Projekt N3 BiLD 
Universit£t des Saarlandes 
Im Stadtwald 15 
D-6600 Saarbrficken 11, FRG 
neumann@coli.uni-sb.de 
ABSTRACT 
A consequent use of reversible grammars within 
natural language generation systems has strong 
implications for the separation into strategic and 
tactical components. A central goal of this paper 
is to make plausible that a uniform architecture 
for grammatical Processing will serve as a basis 
to achieve more flexible and efficient generation 
systems. 
1 Introduction 
In general, the goal of parsing is the derivation of 
all possible grammatical structures defined by a 
grammar of a given string a (i.e. especially the 
determination of all possible logical forms of o~) 
and the goal of the corresponding generation task 
is the computation, of all possible strings defined 
by a grammar of a! given logical form & that are 
logically equivalent to ~ (see also (Shieber, 1988), 
(Calder et al., 1989)). Recently, there is a strong 
tendency to use the same grammar for perform- 
ing both tasks. Besides more practically moti- 
vated reasons - obtaining more compact systems 
or avoidance of inconsistencies between the input 
and output of a system - there are also theoreti- 
cal (a single mode! of language behaviour) and 
psychological evidences (empirical evidence for 
shared processors or facilities, cf. (Garrett, 1982), 
(Frazier, 1982), (Ja'ckendoff, 1987)) to adopt this 
view. 
From a formal point of view the main interest in 
obtaining non-directional grammars is the spec- 
ification of the relationship between strings and 
logical forms. 1 According to van Noord (1990), 
a grammar is reversible if the parsing and gen- 
eration problem is computable and the relation 
between strings and logical forms is symmetric. 
In this case parsing and generation are viewed as 
mutually inverse processes. 
Furthermore there are also approaches that as- 
sume that it is possible to use the same algo- 
rithm for processing the grammar in both direc- 
tions (e.g. (Hasida and Isizaki, 1987), (Shieber, 
1988), (Dymetman et aL, 1990), (Emele and Za- 
jac, 1990)). A great advantage of a uniform pro- 
cess is that a discourse and task independent 
module for grammatical processing is available. 
This means that during performing both tasks 
the same grammatical power is potentially dis- 
posable (regardless of the actual language use). 
Nevertheless, in most of the 'real' generation 
systems where all aspects of the generation pro- 
cess of natural language utterances are consid- 
ered, grammars are used that are especially de- 
signed for generation purposes (cf. (Hovy, 1987), 
(Dale, 1990), (Horacek, 1990), (McKeown el al., 
1990), (Reithinger, 1991)). ~ 
The purpose of this paper is to show that the 
use of a uniform architecture for grammatical pro- 
cessing has important influences for the whole 
generation task. A consequent use of a uniform 
process within a natural language generation sys- 
tem affects the separation into strategic and tacti- 
11 assume a notion of grammars that integrate phono- 
logical, syntactical and semantical levels of description, 
e.g., (Pollard and Sag, 1987). 
2But it is important to note here, that most of the 
proposed grammars are unification-based which is an im- portant common property with 
respect to current parsing 
granmaars. 
31 
cal components. On the one hand, existing prob- 
lems with this separation emerge, on the other 
hand uniform architectures will serve as an im- 
portant (linguistic) basis to achieve first solutions 
for the problems. 
In the next section I will discuss important 
problems and restrictions with the modular de- 
sign of current generation systems and will then 
show why a uniform architecture as the gram- 
matical basis can contribute to solutions of the 
problems. 
2 Modularity in Generation 
Systems 
The Problem It is widely accepted to cut 
down the problem of natural language generation 
(NLG) into two subtasks: 
• determination of the content of an utterance 
• determination of its linguistic realization 
This 'divide and conquer' view of generation is 
the base of current architectures of systems. With 
few exceptions (e.g., (Appelt, 1985)) the following 
two components are assumed: 
• 'what to say' part (strategic component) 
• 'how to say it' part (tactical component) 
But, as it has been demonstrated by some au- 
thors ((hppelt, 1985), (Hovy, 1987), (P~ubinoff, 
1988), (Neumann, 1991), (Reithinger, 1991))it 
is not possible to separate the two phases of the 
generation process completely, e.g., in the case 
of lexieal gaps, choice between near synonyms or 
paraphrases. 
Currently, in systems where the separation is 
advocated the problems are sometimes 'solved' 
in such a way that the strategic component has 
to provide all information needed by the tactical 
component to make decisions about lexical and 
syntactic choices (McDonald, 1983), (McKeown, 
1985), (Busemann, 1990), (Horacek, 1990). As a 
consequence, this implies that the input for tac- 
tical components is tailored to determine a good 
sentence, making the use of powerful grammatical 
processes redundant. In such approaches, tactical 
components are only front-ends and the strategic 
component needs detailed information about the 
language to use. 
Hence, they are not separate modules because 
this implies that both components share the 
grammar. As pointed out in Fodor (1983) one of 
the characteristic properties of a module is that 
it is computationally autonomous. But a rele- 
vant consideration of computationally autonomy 
is that modules do not share sources (in our case 
the grammar). 
Looking for More Symmetric Architec- 
tures To maintain the modular design a more 
symmetric division into strategic and tactical sep- 
aration is needed: 
• Strategic component ~ primarly concerned 
with conceptual decisions 
• Tactical component • , primarly concerned 
with linguistic decisions 
A consequence of this view is that the strate- 
gic component has no detailed information about 
the specific grammar and lexicon. This means 
that in general a message which is constructed 
precisely enough to satisfy the strategic compo- 
nent's goal can be underspecified from the tactical 
viewpoint. For example, if the strategic compo- 
nent specifies as input to the tactical component 
that 'Peter loves Maria', and 'Maria' is the cur- 
rent focus, then in German it is possible to utter: 
1 Maria wird von Peter geliebt 
'Maria is loved by Peter' 
Or 
2 Maria liebt Peter 
'Maria, Peter loves' 
Of course, a 'real' generation system needs to 
choose between the possible paraphrases. An 
adequate generation system should avoid to ut- 
ter 2 because for this utterance there exists also 
the unmarked reading that 'Maria loves Peter'. 
As long as the strategic component has no de- 
tailed knowledge of a specific grammar it could 
not express 'choose the passive form to avoid am- 
biguity'. But then the process can only choose 
randomly between paraphrases during generation 
and this means that the underlying message will 
possibly not be conveyed. 
There is also psychologically grounded evidence 
to assume that the input to a tactical component 
might not be necessary and sufficient to make lin- 
guistic decisions. This is best observed in exam- 
32 
ples of self-correction (Levelt, 1989). For exam- 
ple, in the following utterance: a 
"but aaa, bands like aaa- aaa- aaa- errr- 
like groups, pot bands, - groups, you 
know what I mean like aaa." 
the speaker discovers two words (the near- 
synonymous 'groulp' and 'band') each of which 
comes close to the underlying concept and has 
problems to decide which one is the most suit- 
able. In this case, the problem is because of a 
mis-match between what the strategic compo- 
nent want to express and what the language is 
capable to express (Rubinoff, 1988). 
Current Approaches In order to be able to 
handle these proble~ ins, more flexible tactical com- 
ponents are necessary that are able to handle e.g. 
underspecified inpht. In (Hovy, 1987), (Finkler 
and Neumann, 1989) and (Reithinger, 1991) ap- 
proaches are described how such more flexible 
components can be achieved. A major point of 
these systems is to assume a bidirectional flow 
of control betweenl the strategic and the tactical 
components. 
The problem with systems where a high degree 
of feedback between the strategic and the tactical 
components is necessary in order to perform the 
generation task is that one component could not 
perform its specific task without the help of the 
other. But when the mode of operation of e.g. 
the tactical component is continuously influenced 
by feedback from the strategic component then 
the tactical component will lose its autonomy and 
consequently this means that it is not a module 
(see also (LeveR, 1989)). 
3 Integration of Parsing and 
Generation 
A promising approach for achieving more au- 
tonomous tactical components is to integrate gen- 
eration and parsing in a more strict way. By this 
I mean: 
• the use of resulting structures of one direc- 
tion directly in the other direction, 
aThis example is taken from Rubinoff (1988) and is 
originally from a corpus of speech collected at the Univer- 
sity of Pennsylvania. ' 
• the use of one mode of operation (e.g., pars- 
ing) for monitoring the other mode (e.g., gen- 
eration). 
A main thesis of this paper is that the best 
way to achieve such integrated approach is to use 
a uniform grammatical process as the linguistic 
basis of a tactical component. 
Use of Same Structures in Both Directions 
If parsing and generation share the same data (i.e. 
grammar and lexicon) then it is possible that re- 
sulting structures of one direction could be used 
directly in the other direction. For example, dur- 
ing the generation of paraphrases of the ambigu- 
ous utterance 'Remove the folder with the sys- 
tem tools.' the generator can use directly the 
analysed structures of the two NPs 'the folder' 
and 'the system tools' in the corresponding para- 
phrases. In a similiar way parsing and generation 
of elliptic utterances can also be performed more 
efficiently. For example, consider the following 
dialog between person A and B: 
A: 'Peter comes to the party tonight.' 
B: 'Mary, too.' 
In order to be able to parse B's utterance A can 
directly use parts of the grammatical structure of 
his own utterance in order to complete the elliptic 
structure. 4 
Adaptability to Language Use of Others 
Another very important argument for the use of 
uniform knowledge sources is the possibility to 
model the assumption that during communica- 
tion the use of language of one interlocutor is 
influenced by means of the language use of the 
others. 
For example, in a uniform lexicon it does not 
matter wether the lexeme was accessed during 
parsing or generation. This means that the use of 
linguistic elements of the interlocutor influences 
the choice of lexical material during generation 
if the frequency of lexemes will serve as a deci- 
sion criterion. This will help to choose between 
lexemes which are synonymous in the actual situ- 
ation or when the semantic input cannot be suffi- 
ciently specified. E.g. some drinking-devices can 
be denoted either 'cup' or 'mug' because their 
4In this particular case, A can use the whole VP 'will 
come to the party'. In general the process is more compli- 
cated e.g., if B's answer would be 'Mary and John, too'. 
33 
shape cannot be interpreted unequivocally. An 
appropriate choice would be to use the same lex- 
eme that was previously used by the hearer (if no 
other information is available), in order to ensure 
that the same object will be denoted. In prin- 
ciple this is also possible for the choice between 
alternative syntactic structures. 
This adaptability to the use of language of part- 
ners in communication is one of the sources for 
the fact that the global generation process of hu- 
mans is flexible and efficient. Of course, adapt- 
ability is also a kind of co-operative behaviour. 
This is necessary if new ideas have to be expressed 
for which no mutually known linguistic terms ex- 
ist (e.g., during communication between experts 
and novices). In this case adaptability to the use 
of language of the hearer is necessary in order 
to make possible that the hearer will be able to 
understand the new information. 
In principle this kind of adaptability means 
that the structures of the input computed during 
the understanding process carry some informa- 
tion that can be used to parametrize the genera- 
tion process. This leads to more flexibility: not 
all necessary parameters need to be specified in 
the input of a generator because decision points 
can also be set during run-time. 
This dynamic behaviour of a generation system 
will increase efficiency, too. As McDonald et al. 
(1987) define, one generator design is more effi- 
cient than another, if it is able to solve the same 
problem with fewer steps. They argue that"the 
key element governing the difficulty of utterance 
production is the degree of familiarity with the 
situation". The efficiency of the generation pro- 
cess depends on the competence and experience 
one has acquired for a particular situation. But 
to have experience in the use of linguistic objects 
that are adequate in a particular situation means 
to be adaptable. 
Monitoring As Levelt (1989) pointed out 
"speakers monitor what they are saying and how 
they are saying it", i.e. they monitor not only for 
meaning but also for linguistic well-formedness. 
To be able to monitor what one is saying is very 
important for processing of underspecified input 
and hence for achieving a more balanced divison 
of the generation task (see sec. 2). For exam- 
ple to choose between the two paraphrases of the 
example in sec. 2, the tactical component could 
parse the resulting strings in order to decide to 
choose the less ambiguous string 'Mary is loved 
by Peter.' It only needs to know from the strate- 
gic component that unambiguous utterances are 
preferred (as a pragmactic constraint). 
In Levelt's model parsing and generation are 
performed in an isolated way by means of two 
different grammars. In such flow of control the 
complete structure has to be generated again if 
ambiguities are detected that have to be avoided. 
If, for example an intelligent help-system that 
supports a user by using an operation research 
system (e.g. Unix, (Wilensky et al., 1984)), re- 
ceives as input the utterance "Remove the folder 
with the system tools" then the system is not able 
to perform the corresponding action directly be- 
cause it is ambiguous. But the system could ask 
the user "Do you mean 'Remove the folder by 
means of the system tools' or 'Remove the folder 
that contains the system tools' ". This situation 
is summarized in the following figure (LF' and 
LF" symbolize two readings of S): 
LF' LF" 
S: Remove the folder with 
the system tools 
S~: Remove the folder by means of 
the system tools 
S": Remove the folder that contains 
the system tools 
Figure 1: Relationship between ambiguities and 
paraphrases 
If parsing and generation are performed in an 
isolated way then generation of paraphrases can 
be very inefficient, because the source of the am- 
biguous utterance S is not used directly to guide 
the generation process. 
Generation of Paraphrases In order to clar- 
ify, why an integrated approach can help to solve 
the problem I will consider the problem of gener- 
ation of paraphrases in more detail. 
If a reversible grammar is used in both direc- 
tions then the links between the strings and logi- 
34 
phon : (remove the folder with the system tools) 
synsem : S\[imp\] 
head: synsem : VP\[fin\] 
\[ phon : (the foider) \] dtrs : eomp : ( synsem : N P\[ace\] ) 
adjunct : PP\[< with the system tools >\] 
Figure 2: 'with the system tools' in modifier position of the VP 
" phon : (remove 
synsem S\[imp\] 
!head : \[ 
dtrs : .comp : ( 
the folder with the system tools) 
phon : (remove) 
synsem : VP\[fin\] \] 
phon : (the folder with the system tools) 
synsem : N P\[aec\] 
\[ head:NP\[< the folder >\] \] 
dtrs : adjunct : PP\[< with the system tools >\] 
Figure 3: The same PP as a modifier of the NP 'the folder' 
cal forms in fig. 1 are bidirectional. 5 
A first naive algorithm that performs genera- 
tion of paraphrasds using a reversible grammar 
can be described as follows: Suppose S is the in- 
put for the parser :then the set 
{(S,!LF'), (S, LF")} 
is computed. Now LF' respectively LF" is given 
as input to the generator to compute possible 
paraphrases. The sets 
{(LF', S'), (LF', S)) 
respectively 
{(LF", S), (LF", S")} 
result. By means of comparison of the elements 
of the sets obtained during generation with the 
5It is not of central role here wether the 'competence' 
grammar is actually compiled in two efficient parsing and 
generation grammars ~ long as the symmetry property is 
not affected. This inh~erent property of a reversible gram- 
mar is very important in the case of generation of para- 
phrases because it ensures that the ambiguous structure 
and the corresponding paxaphrases are related together. 
If this would not be ~he case then this would mean that 
one is only able to generate the paraphrases but not the 
ambiguous structure. 
set obtained during parsing one can easily deter- 
mine the two paraphrases S' and S" because of 
the relationship between strings and logical forms 
defined by the grammar. 
This algorithm is naive because of the assump- 
tion that it is possible to generate all possi- 
ble paraphrases at once. Although 'all-parses' 
algorithms are widley used during parsing in 
natural language systems a corresponding 'all- 
paraphrases' strategy is not practicle because in 
general the search space during generation is 
much larger (which is a consequence of the mod- 
ular design discussed in sec. 2). 
Of course, from a formal point of view one is 
interested in algorithms that compute all gram- 
matically well-formed structures - at least poten- 
tially. So most of the currently developed gener- 
ators and uniform algorithms assume - more or 
less explictly - an all-paraphrases strategy (e.g., 
(Shieber, 1988), (Calder et al., 1989), (Shieber et 
al., 1989), (Dymetman et al., 1990), (Emele and 
Zajac, 1990)). But from a practical point of view 
they are not directly usable in such specific situ- 
ations. 
35 
More Suitable Strategies A more suitable 
strategy would be to generate only one para- 
phrase for each ambiguous logical form. As long 
as parsing and generation axe performed in an iso- 
lated way the problem with this strategy is that 
there is no control over the choice of paraphrases. 
In order to make clear this point I will look closer 
to the underlying structure of the example's ut- 
terances. 
The problem why there are two readings is that 
the PP 'with the system folder' can be attached 
into modifier position of the NP 'the folder' (ex- 
pressing the semantic relation that 'folder' con- 
tains 'system tools') or of the verb 'remove' (ex- 
pressing semantically that 'folder' is the instru- 
ment of the described situation). Fig. 2 and 3 
(see above) show the internal grammatical struc- 
ture in a HPSG-style notation (omitting details, 
that are not relevant in this context). 
As long as the source of the ambiguity is not 
known it is possible to generate in both cases the 
utterance 'Remove the folder with the system- 
tools' as a paraphrase of itself. Of course, it is 
possible to compare the resulting strings with the 
input string S. But because the source of the am- 
biguity is not known the loop between the iso- 
lated processes must be performed several times 
in general. 
A better strategy would be to recognize rele- 
vant sources of ambiguities during parsing and 
to use this information to guide the generation 
process. Meteer and Shaked (1988) propose an 
approach where during the repeated parse of an 
ambiguous utterance potential sources of ambigu- 
ity can be detected. For example when in the case 
of lexical ambiguity a noun can be associated to 
two semantic classes a so called 'lexical ambigu- 
ity specialist' records the noun as the ambiguity 
source and the two different classes. These two 
classes are then explicitly used in the generator 
input and are realized as e.g. modifiers for the 
ambiguous noun. 
The only common knowledge source for the 
paraphraser is a high order intensional logic lan- 
guage called World Model Language. It serves as 
the interface between parser and generator. The 
problem with this approach is that parsing and 
generation are performed in an isolated way using 
two different grammars. If an ambiguous utter- 
ance S need to be paraphrased S has to be parsed 
again. During this repeated parse all potential 
ambiguities have to be recognized and recorded 
(i.e. have to be monitored) by means of different 
'ambiguity specialists'. The problem here is that 
also local ambiguities have to be considered that 
are not relevant for the whole structure. 
An Alternative Approach I will now de- 
scribe the basic idea of an approach that is based 
on an integrated approach where both tasks share 
the same grammar. The advantage of this ap- 
proach is that no repeated parse is necessary to 
compute potential ambiguity sources because the 
different grammatical structures determined dur- 
ing parsing are used directly to guide the gen- 
eration process. By means of this strategy it is 
also ensured that an ambiguous utterance is not 
generated as a paraphrase of itself. 
In principle the algorithm works as follows: 
During the generation of paraphrases the gen- 
eration process is monitored in such a way that 
the monitor compares in each step the resulting 
structures of the generation process with the cor- 
responding structures from parsing maintained in 
the alternative parse trees (I will now assume that 
two parse trees P1 and P2 corresponding to the 
structures given in fig. 2 and 3 are obtained dur- 
ing parsing). Suppose that LF' (cf. fig. 1) is 
specified as the input to the generator. In the case 
where the generator encounters alternative gram- 
matical structures to be expanded, the monitor 
guides the generator by means of inspection of 
the corresponding parse trees. In the case where 
actual considered parts pl and p2 of P1 and P2 
(e.g., same NPs) axe equal the generator has to 
choose the same grammatical structure that was 
used to build Pl and p~ (or more efficiently the 
generator can use the partial structure directly as 
a kind of compiled knowledge). In the case where 
a partial structure of e.g. parse tree P1 has no 
correspondence in P2 (cf. fig. 2 and 3) an ambi- 
guity source is detected. In this case an alterna- 
tive grammatical structure has to be chosen. 6 
At this point it should be clear that the easiest 
way in order to be able to generate 'along parsed 
structures' is to use the same grammar in both 
directions. In this case grammatical structures 
obtained during parsing can be used directly to 
restrict the potential search space during genera- 
tion. 
°Of course, the described algorithm is too restrictive, 
in order to handle non-structural (e.g. contextual) para- 
phrases. But, I assume that this approach is also appli- 
cable in the case of lexiccal amibiguities prerequisite word 
meanings are structurally described by means of lexical se- 
mantics (e.g., Jackendoff's Lexiccal Conceptual Structures 
( Jackendoff, 1990)) 
36 
This approach :is not only restricted in cases 
where the input is ambiguous and the para- 
phrases must contrast the different meanings. It 
can also be used for self-monitoring when it has 
to be checked wel~her a produced utterance S of 
an input form LF is ambiguous. In this case S will 
be parsed. If during parsing e.g., two readings LF 
t I . . and LF are deduced LF IS generated again along 
the parse tree obtained for S. Now an utterance 
S' can be generated that has the same meaning 
but differs with respect to the ambiguity source 
of S. 
4 Current Work 
We have now started a project called BiLD (short 
for Bidirectional iLinguistic Deduction) at the 
University of Saarland (Saarbriicken) where it 
will be investigated how an integrated approach 
k of parsing and generation can be realized effi- 
ciently by means of a uniform architecture and 
how such a model can be used for increasing flex- 
ibility and efficiency during natural language pro- 
cessing. 
The main topic lof the BiLD project is the de- 
velopment of a uniform parametrized deduction 
process for grammahcal processing. This process 
constitutes the core of a flexible and symetric tac- 
tical module. In order to realize the integrated 
approach and to obtain a high degree of efficiency 
in both directions'we will develop methods for a 
declarative encoding of information of control in 
the lexicon and grammar. 
We follow a sign-based approach for the de- 
scription of linguistic entities based on Head- 
driven Phrase Structure Grammar (Pollard and 
Sag, 1987) and the variant described in Reape 
(1989). Besides theoretical reasons there are also 
reasons with respect to system's design criterions 
to adopt this view because all levels of descrip- 
tions (i.e. phonological, syntactic and semantic 
structure) of lingffistics entities (i.e. words and 
phrases) are described simultanueous in a uni- 
form way by means of partial information struc- 
tures. None of the levels of description has a 
privileged status but expresses possible mutually 
co-ocurrence restrictions of structures of different 
levels. 
Furthermore a high degree of lexicalism is as- 
sumed so that the lexicon as a complex hierachi- 
cal ordered data Structure plays a central role 
in BiLD. As it has been shown this lexicalized 
view supports revei'sibility (el. (Newman, 1990), 
(Dymetman et al., 1990)) and the performing 
of specific processing strategies (e.g., incremental 
and parallel generation, (Neumann and Finkler, 
1990)). 
The task of the deduction process during gener- 
ation is to construct the graphemic form of a spec- 
ified feature description of a semantic form. For 
example, to yield the utterance "A man sings." 
the deduction process gets as input the semantic 
feature structure 
tel : sing' 
agens: 
quant : exists' 
oar : 
restr : \[ pred : man' \] var : 
and deduces the graphematic structure 
\[ graph : (A_man_sings.) \] 
by means of successive application of lexical and 
grammatical information. In the same way the 
deduction process computes from the graphe- 
matic structure an appropriate semantic struc- 
ture in parsing direction. A first prototype based 
on head-driven bottom-up strategy is now under 
development (cf. (van Noord, 1991)). 
A major aspect of the BiLD project is that 
a specific parametrization of the deduction pro- 
cess is represented in the lexicon as well as in the 
grammar to obtain efficient structures of control 
(Uszkoreit, 1991). The main idea is that pref- 
erence values are assigned to the elements (dis- 
juncts or conjuncts) of feature descriptions. For 
example, in HPSG all lexical entries are put to- 
gether into one large disjunctive form. From a 
purely declarative point of view these elements 
are unordered. But a preference structure is used 
during processing in order to guide the process 
of lexical choice efficiently which itself influences 
the grammatical process. 
5 Conclusion 
A main thesis of this paper was to show that ex- 
isting problems with the modular design of cur- 
rent generation systems emerge when a reversible 
grammar is used. In order to maintain the mod- 
ular design I have proposed an approach that 
is based on a strong integration of parsing and 
generation of grammatical structures using a re- 
versible grammar and monitoring mechanisms. 
37 
By means of such an integrated approach per- 
forming of e.g. generation of paraphrases can be 
done more easier and efficently. 
Acknowledgements 
This research was supported by SFB 314, Project 
N3 BiLD. 

Bibliography 
Douglas E. Appelt. Planning English Sentences. 
Cambridge University Press, Cambridge, 1985. 
Stefan Busemann. Generierung natlirlicher 
Sprache mit Generalisierten Phrasenstruktur- 
Grammatiken. PhD thesis, University of Saar- 
land (Saarbriicken), 1990. 
Jonathan Calder, Mike Reape, and Henk Zeevat. 
An algorithm for generation in unification cat- 
egorial grammar. In Fourth Conference of the 
European Chapter of the Association for Com- 
putational Linguistics, pages 233-240, Manch- 
ester, 1989. 
Robert Dale. Generating receipes: An overview 
of epicure. In Robert Dale, Chris Mellish, 
and Michael Zock, editors, Current Research in 
Natural Language Generation, pages 229-255. 
Academic Press, London, 1990. 
Mare Dymetman, Pierre Isabelle, and Francois 
Perrault. A symmetrical approach to parsing 
and generation. In Proceedings of the 13th In- 
ternational Conference on Computational Lin- 
guistics (COLING), pages 90-96, Helsinki, 
1990. 
Martin Emele and Po!mi Zajac. Typed unification 
grammars. In Proceedings of the 13th Interna- 
tional Conference on Computational Linguis- 
tics (COLING), pages 293-298, Helsinki, 1990. 
Wolfgang Finkler and Giinter Neumann. Popel- 
how: A distributed parallel model for incre- 
mental natural language production with feed- 
back. In Proceedings of the Eleventh Inter- 
national Joint Conference on Artificial Intel- 
ligence, pages 1518-1523, Detroit, 1989. 
Jerry A. Fodor. The Modularity of Mind: An Es- 
say on Faculty Psychology. A Bradford Book, 
MIT Press, Cambridge, Massachusetts, 1983. 
Lyn Frazier. Shared components of production 
and perception. In M. A. Arbib et al., editor, 
Neural Models of Language Processes, pages 
225-236. Academic Press, New York, 1982. 
Merrill F. Garrett. Remarks on the relation be- 
tween language production and language com- 
prehension systems. In M. A. Arbib et al., 
editor, Neural Models of Language Processes, 
pages 209-224. Academic Press, New York, 
1982. 
K. Hssida and S. Isizaki. Dependency propa- 
gation: A unified theory of sentence compre- 
hension and generation. In Proceedings of the 
Tenth International Joint Conference on Artifi- 
cial Intelligence, pages 664-670, Mailand, 1987. 
Helmut Horacek. The architecture of a genera- 
tion component in a complete natural language 
dialogue system. In Robert Dale, Chris Mel- 
lish, and Michael Zock, editors, Current Re- 
search in Natural Language Generation, pages 
193 - 227. Academic Press, London, 1990. 
Eduard. H. Hovy. Generating Natural Language 
under Pragmatic Constraints. PhD thesis, Yale 
University, 1987. 
Ray Jackendoff. Consciousness and the Compu- 
tational Mind. MIT Press, Cambridge, Mas- 
sachusetts, 1987. 
Ray Jackendoff. Semantic Structures. MIT 
Press, Cambridge, Massachusetts, 1990. 
Willem J. M. Levelt. Speaking: From Intention 
to Articulation. MIT Press, Cambridge, Mas- 
sachusetts, 1989. 
David D. McDonald. Natural language genera- 
tion as a computational problem: An intro- 
duction. In M. Brady and C. Berwiek, edi- 
tors, Computational Models of Discourse. MIT 
Press, Cambridge, Massachusetts, 1983. 
David D. McDonald, Marie W. Meteer, and 
James D. Pustejovsky. Factors contributing to 
efficiency in natural language generation. In 
K. Kempen, editor, Natural Language Gener- 
ation: New Results in Artificial Intelligence, 
Psychology and Linguistics, pages 159-182. 
Martinus Nijhoff, Dordrecht, 1987. 
Kathleen R. McKeown. Text Generation: Using 
Discourse Strategies and Focus Constraints to 
Generate Natural Language Text. Cambridge 
University Press, Cambridge, 1985. 
Kathleen R. McKeown, Michael Elhadad, Yu- 
miko Fukomoto, Jong Lim, Christine Lom- 
bardi, Jacques Robin, and Frank Smadja. Nat- 
ural language generation in comet. In Robert 
Dale, Chris Mellish, and Michael Zock, editors, 
Current Research in Natural Language Genera- 
tion, pages 103 2 139. Academic Press, London, 
1990. 
Marie Meteer and Varda Shake& Strategies for 
effective paraphrasing. In Proceedings of the 
12th International Conference on Computa- 
tional Linguistics (COLING), Budapest, 1988. 
Gfinter Neumann. A bidirectional model for nat- 
ural language processing. In Fifth Conference 
of the Europea'n Chapter of the Association 
for Computational Linguistics, pages 245-250, 
Berlin, 1991. 
Gfinter Neumann and Wolfgang Finkler. A head- 
driven approach to incremental and parallel 
generation of syntactic structures. In Proceed- 
ings of the 13th International Conference on 
Computational Linguistics (COLING), pages 
288-293, Helsinki, 1990. 
Paula Newman. Towards convenient bi- 
directional grammar formalisms. In Proceed- 
ings of the 13th International Conference on 
Computational iLinguistics (COLING), pages 
294-298, Helsinki , 1990. 
Carl Pollard and ivan Sag. Information Based 
Syntax and Semantics, Volume 1. Center for 
the Study of Language and Information Stan- 
ford, 1987. 
Mike Reape. A logical treatment of semi-free 
word order and bounded discontinuous con- 
stituency. In FOurth Conference of the Euro- 
pean Chapter of the Association for Computa- 
tional Linguistics , pages 103-110, Manchester, 
1989. 
Norbert Reithinger. Popel: A parallel and in- 
cremental natural language generation system. 
In C. L. Paris et al., editor, Natural Language 
Generation in Artificial Intelligence and Com- 
putational Linguistics, pages 179-199. Kluwer, 
1991. 
Robert Rubinoff. A cooperative model of strat- 
egy and tactics in generation. In Paper 
presented at the' Fourth International Work- 
shop on Natural: Language Generation, Santa 
Catalina Island, 1988. 
Stuart M. Shieber. A uniform architecture for 
parsing and generation. In Proceedings of the 
12th International Conference on Computa- 
tional Linguistics (COLING), Budapest, 1988. 
Stuart M. Shieber, Gertjan van Noord, Robert C. 
Moore, and Fernando C.N. Pereira. A 
semantic-head-driven generation algorithm for 
unification based formalisms. In 27th Annual 
Meeting of the Association for Computational 
Linguistics, Vancouver, 1989. 
Hans Uszkoreit. Strategies for adding control in- 
formation to declarative grammars. In 29th 
Annual Meeting of the Association for Com- 
putational Linguistics, Berkeley, 1991. 
Gertjan van Noord. Reversible unification-based 
machine translation. In Proceedings of the 
13th International Conference on Computa- 
tional Linguistics (COLING), Helsinki, 1990. 
Gertjan van Noord. Towards uniform process- 
ing for constraint-based categorial grammars. 
In Proceedings of the ACL Workshop on Re- 
versible Grammar in Natural Language Pro- 
cessing, Berkeley, 1991. 
R. Wilensky, Y. Arens, and D. Chin. Talking to 
unix in english: An overview of uc. Communi- 
cations of the ACM, pages 574 -593, 1984. 
