A New Model for Lexical Choice for Open-Class Words 
Ehud Reiter'~ 
Aiken Computation Laboratory 
Harvard University 
Cambridge, Mass 02138 
Abstract 
The lexical choice process should be regarded as a con- 
straint satisfaction problem: the generation system must 
choose a lexical unit that is accurate (t~mthful), va//d 
(conveys the necessary information), and preferred (max- 
irnal under a preference function). This corts~aint-based 
architecture allows a clema separation to be made 
between what the system knows of the object or event, 
and what the system wishes to communicate about the 
object or event. It also allows lexical choices to be 
biased towards basic-level (Rosch 1978) and other pre- 
ferred lexical units. 
1. Introduction 
Lexical choice for open-class words has typically been 
regarded as a matching or classification problem. The 
generation system is given a semantic structure that 
represents an object or event, and a dictionary that 
represents the semantic meanings of the lexical units 
(Zgusta 1971) of the target language; it then chooses the 
lexical unit (or set of lexical units) that best matches the 
object or event. This paper proposes an alternative lexi- 
cal choice architecture, in which the lexical choice pro- 
cess is regarded as a constraint satisfaction problem: the 
generation system must choose a lexical unit that is 
accurate (truthful), valid (conveys the necessary infor- 
marion), and preferred (maximal under a preference 
function). 1 This constraint-based architecture is more 
robust than classification systems. In parricular, it 
allows a clean separation to be made between what the 
system knows of the object or event, and what the sys- 
tem wishes to communicate about the object or event; 
and it allows lexical choices to be biased towards 
basic-level (Rosch 1978) and other preferred lexical 
units. 
Throughout this paper, it will be assumed that both 
lexical units and objects/events are represented as 
t Currently at the Department of Artificial Intelligence, 
University of Edinburgh, 80 South Bridge, Edinburgh EH1 1HN, 
Scotland. E-maih reitel@aitma.edinburgh.ac.uk 
I This paper does not exmnine the kind of oollocational and 
selectional constraints discussed by Cumming (1986) and Niren- 
burg and Nirenburg (1988). 
classes in a KL-ONE type taxonomy (Brachman and 
Schmolze 1985). For example, the lexical unit 
Bachelor might be represented as the generic class 
(Human with role value restrictions Sex:Male, Age- 
status:Adult, Married:False); and the object Terry might 
be represented as the individual class (Human with role 
fillers Sex:Male, Eye-color:Brown, Birthplace:Chicago, 
Employer:IBM .... ). Default attributes as well as 
definitional information can be associated with lexical 
units; this is essential for making appropriate lexical 
choices (Section 5). Figure 1 shows a sample taxonomy 
that will be used for most of the examples in this paper. 
Lexical units (e.g., Bachelor) are shown in bold font, 
while objects (e.g., Terry) are shown in italic font. 
Role value restrictions (VR's), such as Sex:Male for 
Man, are fisted textually instead of displayed graphi- 
cally, to simpfify the complexity of the diagram; default 
attributes (e.g., Can-fly:True for Bird) are listed in italic 
font. Basic-level classes (e.g., Man) are underlined. 
Section 2 of the paper discusses classification-based 
systems and some of the problems associated with them. 
Section 3 introduces the proposed constraint-based sys- 
tem; Section 4 looks in more detail at the lexical prefer- 
ences used by the system; and Section 5 briefly 
discusses the need for default atlributes in the semantic 
representations of lexical units. The constraint-based 
lexical choice system has been incorporated into the FN 
system (Reiter 1990), which generates certain kinds of 
natural language object descriptions. FN uses some 
additional preference rules that primarily affect NP for- 
marion; these rules are not discussed in this paper. 
2. Lexical Choice as Classification 
The two major approaches (to date) for lexical choice 
have been discrimination nets and structure mapping 
systems. Both of these approaches can be regarded as 
classification/matching architectures, where a classifier 
is given an object or event, and is asked to find an 
appropriate lexical unit that fits that object or event. 
Discrimination nets (e.g., Goldman 1975; Pustejovsky 
and Nirenburg 1987) are basically decision trees. They 
are typically used as high-speed 'compiled' classifiers 
that select the most specific lexical unit that subsumes 
23 
C Animal 
Vertebrate 
Objec~ 
Machine 
t Network 
Fish 
t Shark 
(Dangerous:True) ~ 
tL 
Breathes:Air 
Human ) ( Pekingese 
Brca~es.-Air (Can-fly:True) j 
larrow 
Ethernet Dam-rate: lOMbiffsec l 
Circuit-type:Packet \[ Physical-medium:Coaxial-cableJ 
0 
I Adult Age-status:Adult I 
f\ I ~, I \[ ~=ma~o I 
Bachelor Married:False I 
cPrimitive 3 lass 
\[ Defined \[ 
Class 
-.@ 
(Key) 
Basic Level Class defining role VR 
Lexical Unit Class (default role filler) 
Object Class 
is ~y 
caOStrich n-fly:FalseJ 
Figure 1: Objects 
Lexical Units in a 
Taxonomy 
and 
24 
the target object or event. For instance, looking at 
some of Goldman's examples, the event 
!ngest(John,Milk027), which can be represented in KL- 
ONE as (Ingest with VR's actor:John and 
theme:Milk027), has as its most specific subsuming lexi- 
cal unit (Ingest with VR theme:Liquid), and thus is lexi- 
cally realized as "drink". Similarly, the action 
Ingest(BearO36,Fish802), which'can be represented in 
KL-ONE as (Ingest with VR's actor:Bear036 and 
theme:Fish802), has (Ingest with VR's agent:Non- 
human-animal and theme:Solid) as its most specific sub- 
sumer in a taxonomy of German lexical units, and thus 
is realized, in German, as "fressen". 
Structure-mapping systems (e.g., Jacobs 1987; Iordan- 
skaja et al. 1988; note that different terminology is 
used in different papers) take as input a semantic struc- 
ture that needs to be communicated to the user, search 
for pieces of the input structure that are equivalent to 
lexical units, and then replace the matched structure by 
the corresponding lexical unit. The matching and sub- 
sdtution process continues until the semantic structure 
has been completely reformulated in terms of lexical 
units. For example, the structure (Human (:sex male) 
(:age-status adult) (:wealth high)) might be mapped into 
the structure ("man" (:attribute "rich")), and hence lexi- 
tally realiTcd as "rich man". In KL-ONE terms, the 
matching process can be considered to be a search for a 
class definition that uses only classes and role VR's that 
can be realized as lexical units; e.g., the above example 
essentially redefines the class (Human with role VR's 
Sex:Male, Age-status:Adult, Wealth:High) as the 
equivalent class ("man" with VR "rich"), where the 
lexical unit "man" represents the class (Human with 
role VR's Sex:Male, Age-status:Adult), and the lexical 
unit "rich" is equivalent to the role VR Wealth:High. 
Recently, the machine translation group at CMU has 
proposed an alternative lexical choice system that is 
based on a variant of nearest neighbor classification 
(Center for Machine Translation 1989; Nirenburg et al. 
1987). In the CMU system, both objects and lexical 
units are treated as points or regions in a feature space, 
and the classifier works by choosing the lexical unit that 
is closest to the target object, using a fairly complex 
distance (matching) metric (collocation constraints are 
also taken into consideration). For example, the object 
(Human with VR's Sex:Male and Age:13) would be 
judged closest to the lexical unit (Human with VR's 
Sex:Male and Age:range(2,15)), and thus would be real- 
ized as "boy'. 
All of the above classification-based lexical-ehoice 
architectures 2 suffer from two basic flaws: 
• they do not allow a clean separation to be made 
between what the system knows, and what it wishes 
to communicate; 
• they do not provide a clean mechanism for allowing 
the lexical choice process to be biased towards pre- 
ferred lexical units. 
These failures may lead classification-based systems to 
choose inappropriate lexical units that carry unwanted 
conversational implicatures (Grice 1975), and therefore 
mislead the user. 
2.1. One Input vs Two Inputs 
Classification-based systems take as their input a single 
set of attributes about the object/event being lexicalized, 
and use this set of attributes to select a matching 
classification. However, lexical choice systems should 
look at two input sets of attributes: the set of 
object/event attributes that are relevant and need to be 
conveyed to the user, and the set of attributes that con- 
stitute the system's total knowledge of the object/event 
being lexicalized. 
A lexieal choice system that looks only at the 
system's domain knowledge about the object/event, and 
ignores the set of relevant attributes, may choose inap- 
propriate lexical items that carry unwanted relevance 
conversational implicatures. In particular, a system that 
simply selects the most specific lexical unit that sub- 
sumes the object/event (as many discrimination net sys- 
tems do) may mislead the user by choosing lexical units 
that are too specific. For example, consider the follow- 
ing exchange: 
1) A: "Is Terry a woman?" 
2a) B: "No, Terry is a man" 
2b) B: "No, Terry is a bachelor" 
B's communicative goal is simply to inform A that 
Terry has the attributes {Human, Age-status:Adult, 
Sex:Male}, so utterance (2a) is an appropriate response. 
A lexical choice system that simply selected the most 
specific lexical unit that subsumed Terry would generate 
utterance (2b), however. Utterance (2b) is inappropri- 
ate, and would probably lead A to infer the (incorrect) 
conversational implicature that B thought that Terry's 
marital status was relevant to the conversation. 
A lexical choice system that looks only at the attri- 
butes being communicated, and ignores the system's 
2 Individual lexical-daoice systems can, of course, be aug- 
mented with special code that addresses some of these issues; the 
claim is that the classification-based lexical-choice architeoures 
do not easily or naturally deal with these problems. 
25 
general domain knowledge about the object/event, may 
also make inappropriate lexical choices that lead to 
unwanted conversational implicatures. For example, 
suppose A wished to communicate to B that XNET was 
a Network with the attributes {Data-rate:lOMbit/sec, 
Circuit-type:Packet-switched}. Consider three possible 
lexicalizations: 
3a) "XNET is a network" 
3b) "XNET is a I0 Mbit/sec packet-based network" 
3c) "XNET is an Ethernet" 
Utterance (3c) is the most appropriate utterance (assum- 
ing the user has some domain knowledge about Ether- 
nets). Utterance (3a), however, would be generated by 
a system that simply chose the most specific lexieal unit 
that subsumed {Network, Data-rate:lOMbit/sec, 
Circuit-type:Packet-switched}. 3 This utterance fails to 
fulfill the communicative goal of informing the reader 
that the network has the attributes {Data- 
rate:lOMbitlsec, Circuit-type:Packet-switched}, and is 
therefore unacceptable. Utterance (3b) would be gen- 
erated by a structure-mapping system that chose a lexi- 
cal unit according to the above strategy, and then added 
expficit modifiers to communicate attributes that were 
not impfied by the lexical class. 4 Tlds utterance success- 
fully communicates the relevant information, but it also 
implicates, to the knowledgeable hearer, that XNET is 
not an Ethernet m because if it was, the knowledgeable 
hearer would reason, then the speaker would have used 
utterance (3e). 
2.2. Preferred Lexical Units 
Certain lexical units, in particular those that represent 
basic-level classes (Rosch 1978), are preferred and 
should be chosen whenever possible. Cruse (1977) and 
3 Another possibility is choosing the most general lexical unit 
that is subsumed by the attributes being communicated. Howev- 
er, this cmnot be done by a ~/stma that ignores the object and 
only 1oo1~ at the attributes being communicated, because such a 
system would not know which le~ical units accurately described 
the object. For example, if there were two classes Ethernet and 
Applenet that had the attributes (Network, Data-rate:lOMbitlsec, 
Circuit.type:Packet-switched}, the system could only decide 
whether to generate "Ethemet" or "Applenet" by detexmining 
which of these classes subsumed the object being described (e.g., 
"Etbemet" should be used to describe XNET). See also exam- 
pie 5, where the most appropriate leexical unit that informs the 
hearer that F/do has the attributes {An/ma/, Breathes:Air} is 
"dog', not "mmnmal" or "animal'. 
,t In this example, the 'lexical choice' system is assumed to 
capable of forming a complete NP. In general, it is often 
difficult to separate the task of selecting a single word from the 
task of forming a complete phrase. 
others have suggested that the failure to use a basic- 
level class in an utterance will conversationally impli- 
cate that the basic-level class could not have been used. 
For example, consider the following utterances: 
4) A: "I want to flood room 16 with carbon dioxide" 
5a) B: "Wail there is an animal in the room" 
5b) B: "Wail there is a dog in the room" 
5c) B: "Wait, there is a Pekingese in the room" 
Assume the object in question is Fido, and A's com- 
municative goal is simply to inform B that Fido has the 
attributes {An/ma/, Breathes:Air}, and hence would be 
adversely affected if the room was flooded with carbon 
dioxide. Utterances (5a), (5b), and (5c) all fulfill this 
communicative goal (assuming that Breathes:Air is a 
default attribute of An/ma/), but utterance (5b) is pre- 
ferred because Dog is a basic-level class. Utterance 
(5a) is odd because the use of the superordinate class 
Animal impficates, according to Cruse's hypothesis, that 
the animal in question is not a Dog, Cat, or other com- 
monly known type of animal (or at least the speaker 
does not know that the animal is a member of one of 
these species); utterance (5c) is odd because the use of 
the subordinate class Pekingese implicates that it is 
somehow relevant that the animal is a Pekingese and 
not some other kind of dog. If both of these impfica- 
tures are incorrect, the speaker should choose the lexical 
unit Dog if he wishes to avoid misleading the hearer. 
It should be pointed out that the strategy of simply 
always picking a basic-level class that subsumes the 
object/event will not work, because it ignores the 
system's communicative goals. For instance, a system 
that followed the basic-level strategy would, in the 
situation of example 3, generate utterance (3a) or (3b). 
Both of these are inappropriate and implicate, to the 
knowledgeable user, that utt~ce (3c) could not have 
been used, i.e., that XNET is not an Ethernet. 
3. Lexical Choice as Constraint Satisfaction 
The above problems can be avoided by regarding lexi- 
cal choice as a constraint-satisfaction task instead of a 
classification task. More precisely, the task of choosing 
an appropriate open-class lexical unit should be formal- 
ized as follows: 
Input. 
• Entity: a taxonomy class that represents the system's 
knowledge of the object or event being lexicalized. 
• To-Communicate: a set of predicates (attributes) that 
represent the relevant information about the object 
that needs to be communicated to the user. 
26 
Output: A lexical unit Lex that is a member of the 
knowledge-base taxonomy, and that satisfies the follow- 
ing constraints: 
• Accurate: Lex must be a truthful description of Entity. 
Formally, Lex must subsume Entity. 
• Valid: The use of Lex in an utterance must inform the 
user that the predicates in To-Communicate hold for 
Entity. Formally, every predicate in To-Communicate 
must either be inferrable from the definition of Lex 
(e.g., subsume Lex), or be a default attribute that is 
associated with Lex. 
• Preferred: Lex must be a maximal element of the set 
of accurate and valid lexical units under certain lexi- 
cal preference rules (Section 4). 
In other words, the lexical choice system is given two 
inputs, which represent the system's knowledge of the 
object or event, and the relevant information about that 
object or event that needs to be communicated to the 
user; and is expected to produce as its output a maximal 
lexical unit (under the lexical preference rules) that is 
truthful and conveys the relevant information. 
The constraint-based system makes appropriate lexi- 
cal choices in each of the previous examples: 
• Entity = Terry, To-Communicate = {Human, 
Sex:Male} (example 2). Both Man and Bachelor are 
accurate and valid lexical units. Man is chosen, 
because it is basic-level and therefore preferred. 
• Entity = XNET, To-Communicate = {Network, Data- 
rate :l OMbitlsec, Circuit-type:Packet-switched} 
(example 3). Ethernet is chosen, because it is the 
only accurate and valid lexical unit. 
• Entity = Fido, To-Communicate = {Animal, 
Breathes:Air} (example 5). Accurate and valid lexi- 
cal units include Animal, Mammal, Dog, and Pek- 
ingese. Dog is chosen, because it is basic-level. 
4. Preferences Among Lexical Classes 
If several lexical units are accurate and valid, a set of 
lexical preferences rules is used to select the lexical 
unit the system will utter. The preference for basic- 
level classes was previously mentioned (Section 2.2), 
but it is complicated by entry-level effects (Section 4.1), 
Additional lexical preferences include the length\[subset 
preference (Section 4.2). Combined, the lexical prefer- 
ence rules impose a lexical preference hierarchy on the 
lexical units in the knowledge base. Figure 2 shows 
part of the lexical preference hierarchy that is associated 
with the knowledge base of Figure 1. 
4.1. Basic-Level vs Entry-Level Preferences 
Hirschberg (1985) has suggested that it may be better to 
use Jolicoeur et al.'s (1984) notion of entry level classes 
instead of Rosch's basic level classes. The difference is 
that under the entry-level hypothesis, which category is 
unmarked (i.e., which category may be used without 
generating a conversational implicature) may depend on 
how atypical the object is. For example, consider:. 
(the speaker points to a robin) 
7a) ``Look at the bird" 
7b) "Look at the robin" 
(the speaker points to an ostrich) 
8a) "Look at the bird" 
8b) "Look at the ostrich" 
Under the basic-level hypothesis, a category is either 
basic-level or it is not, and ff it is basic-level, then it is 
always the unmarked way of referring to any object that 
belongs to it. Therefore, under this hypothesis utter- 
ances (7a) and (8a) are both unmarked and carry no 
conversa6onal implicatures, since Bird is a basic-level 
category for most urban Americans. Under the entry- 
level hypothesis, in contrast, while a basic-level 
category is the unmarked way of referring to 'normal' 
members of the category, it may not be the unmarked 
way of referring to atypical members. Instead, a more 
specialized category may be the unmarked way of refer- 
ring to atypical members. Thus, under the entry-level 
hypothesis, even ff utterance (7a) was the unmarked 
way of referring to robins (which are typical birds), 
utterance (8b) could still be the unmarked way of refer- 
ring to ostriches (which are atypical birds). 
The lexical-choice system can allow for entry-level 
effects ff it allows any lexical unit to be marked as 
basic-level in the taxonomy, but then only considers the 
lowest such marked class to be a true basic-level (and 
hence lexically-preferred) class for an object. More 
precisely, ff an object has two subsumers A and B that 
are both marked as basic-level classes, and A subsumes 
B, then the system should only treat B as a lexically- 
preferred class for the object. For example, in Figure 1 
Bird and Ostrich are both marked as basic-level. There- 
fore, the lexical-choiee system should treat Bird (but not 
Sparrow) as a lexically-preferred class for Tweety (a 
Sparrow), and Ostrich (but not Bird) as a lexically- 
preferred class for Big-Bird (an Ostrich). 
4.2. Length/Subset Preferences 
A lexical unit A is almost always preferred over a lexi- 
cal unit B if A's surface form uses a subset of the 
words used by B's surface form (this can be considered 
27 
Fish Dog Man ) ~_Bird ) \] 
x 
~arro~ CPekingese ")C Bachelor) 
hark parrow ~ 
Basic Level Preference Word Subset Preference 
.................. .X ................................................. V 
Figure 2: Some of the Lexical Preferences from Figure 1 
28 
to be a consequence of Grice's maxim of quantity 
(Grice 1975)). Consider, for example, 
9a) ``Don't go swimming; there is a shark in the water" 
9b) "Don't go swimming; there is a tiger shark in the 
water" 
According to the subset lexical, preference rule, lexical 
unit Shark is preferred over lexical unit Tiger-shark. 
Therefore, the use of utterance (9b) carries the conver- 
sational implicature that utterance (9a) could not be 
used, i.e., that it was relevant that the animal was a 
Tiger-shark and not some other kind of Shark. A hearer 
who heard utterance (9b) might infer, for example, that 
the speaker thought that tiger sharks were unusually 
dangerous kinds of sharks. If no such implicature was 
intended by the speaker, then he should use utterance 
(9a), not utterance (9b). 
A stronger version of this preference rule would be to 
prefer lexical unit A to lexical unit B if A's surface 
form used fewer open-class words than B's surface 
form. This would, for example, correctly predict that 
Dog is preferred over Great-Dane, and that Flower is 
preferred over Rocky-Mountain-iris. This preference is 
usually accurate, but it does fail in some cases. For 
example, it is questionable whether Porsche is preferred 
over Sports-car, and doubtful whether Mammal is pre- 
ferred over Great-Dane. 
There are cases where the basic-level preference 
conflicts with (and takes precedence over) both the sub- 
set and the length preferences. Such conflicts are prob- 
ably rare, because psychological and linguistic findings 
suggest that basic-level classes are almost always lexi- 
cally realized with single words (Rosch 1978; Berlin et 
al. 1973). However, there are a small number of basic- 
level classes that have multi-word reali7ations, and this 
can lead to conflicts of the above type. Consider, for 
example, 
10a) "Joe has a mach/ne" 
10b) ``Joe has an appliance" 
10c) "Joe has a washing machine" 
Washing-machine is probably basic-level for most 
Americans. Therefore, utterance (10c) is preferred over 
utterances (10a) and (10b), despite the fact that the 
length preference suggests that utterances (10a) and 
(10b) should be preferred over utterance (10c), and the 
subset preference suggests that utterance (10a) should 
be preferred over utterance (10c). 
4.3. Other Lexical Preference 
There are lexical preferences that are not captured by 
either the basic-level preference or the subset/length 
preference. For example, suppose the speaker wished to 
refer to two animals, a horse and a cow. Consider the 
difference between 
lla) "Look at the an/mats" 
1 lb) "Look at the mamma/s" 
1 lc) "Look at the vertebrates" 
None of the above are basic-level classes (Horse and 
Cow are basic-level for most urban Americans). There- 
fore, neither the basic-level nor the length/subset rules 
indicate any preferences among the above. However, it 
seems clear that utterance (lla) is much preferable to 
utterance (lib), and that utterance (lib) is probably 
preferable to utterance (llc). In addition, the use of 
utterances (llb) or (llc) seems to implicate that utter- 
unee (lla) could not have been used. 
5. Default Attributes 
One final point is that the representation of the seman- 
tics of lexical units must include default attributes as 
well as definitional information. These defaults may 
represent domain knowledge (e.g., birds typically fly) or 
useful conventions that have evolved in a particular 
environment (e.g., most computers at Harvard's Aiken 
Computation Lab run the UNIX operating system). 
Systems that ignore default attributes may make inap- 
propriate lexical choices, and therefore generate utter- 
anoes that carry unwanted conversational implicatures. 
For example, ff To-Communicate was {Bird, Can- 
fly:True}, and Entity was Tweety, consider the 
difference between 
12a) "Tweety is a b/rd" 
12b) ``Tweety is a bird that can fly" 
If the generation system ignored default attributes, it 
would have to generate something like utterance (12b). 
Utterance (12b) sounds odd, however, and a person who 
heard it might infer unwanted and unintended conversa- 
tional implicatm'es, e.g., that some other bird under dis- 
cussion was not able to fly. Utterance (12a) is much 
better, but it can only be generated by a generation sys- 
tem that takes into consideration the fact that Can- 
fly:True is a default attribute of Bird. 
For another example, sutvose an NLG system wished 
to inform a user that a particular computer was a VAX 
that ran the UNIX operating system and the Latex text 
processor (i.e., To-Communicate = {VAX, Operating- 
29 
system:UNIX, Available-software:Latex}). Consider two 
possible utterances: 
13a) "Hucl is a VAX that runs Latex" 
13b) "Hucl is a UNIX VAX that runs Latex" 
Utterance (13a) is acceptable, and indeed expected, if 
the user thinks that Operating-system:UNIX is a default 
attribute of VAX's in the current environment (e.g., at 
the Aiken Computation Lab). In a different environ- 
ment, where users by default associate Operating- 
system:VMS with VAX's, utterance (13a) would be 
misleading and unacceptable, and utterance (13b) should 
be generate& 
6. Conclusion 
This paper has proposed a lexical choice system that 
searches for lexical units that are accurate, valid, and 
preferred with respect to the information the generation 
system wishes to communicate (To-Communicate), and 
the object or event being lexicalized (Entity). This sys- 
tem is more robust than discrimination nets and other 
existing classification-based lexical choice systems, and 
in particular is less likely to make inappropriate lexical 
choices that lead human readers to infer unwanted 
conversational implicatures. The improved performance 
is largely a consequence of the fact that the system 
allows a clean separation to be made between what the 
system knows, and what it wishes to communicate; and 
the fact that the system allows lexical choice to be 
biased towards preferred lexical units. 
Acknowledgements 
Many thanks to Joyce Friedman, Barbara Grosz, Chris 
Mellish, Stuart Shieber, and Bill Woods for their help. 
This work was partially supported by a National Sci- 
ence Foundation Graduate Fellowship, an IBM Graduate 
Fellowship, and a conlract from U S West Advanced 
Technologies. Any opinions, findings, conclusions, or 
recommendations are those of the authors and do not 
necessarily reflect the views of the National Science 
Foundation, IBM, or U S West Advanced Technologies. 

References 
Berlin, B.;Brecdlove, D.; and Raven, P. 1973 General 
Principles of Classification and Nomenclature in Folk 
Biology. American Anthropologist 75:214-242. 
Brachman, R. and Schmolze, J. 1985 An overview of 
the KL-ONE knowledge representation system. Cog- 
nitive Science 9:171-216. 
Center for Machine Translation 1989 KBMT-89 Project 
Report. Carnegie-Mellon University. 
Cruse, D. 1977 The pragmatics of lexical specificity. 
Journal of Linguistics 13:153-164. 
Cumming, S. 1986 The Lexicon in Text Generation. ISI 
Research Report ISI/RR-86-168. Information Sciences 
Institute, University of Southern California. 
Goldman, N. 1975 Conceptual generation. In R. Schank 
and C. Riesbeck (Eds.), Conceptual Information Pro- 
cessing~ American Elselvier. New York. 
Grice, H. 1975 Logic and conversation. In P. Cole and 
J. Morgan (Eds.), Syntax and Semantics: Vol 3, 
Speech Acts, pg 43-58. Academic Press: New York. 
Hirschberg, J. 1985 A Theory of Scalar Implicature. 
Ph.D thesis. Report MS-CIS-85-56, LINC LAB 21, 
Department of Computer and Information Science, 
University of Pennsylvania. 
Iordanskaja, L.; Kittredge, R.; Polguere, A. 1988 Imple- 
menting a Meaning-Text Model for Language Gen- 
eration. Presented at COLING 1988 (not in proc.). 
Jacobs, P. 1987 Knowledge-Intensive Natural Language 
Generation. Artificial Intelligence 33:325-378. 
Jolicoeur, P.; Gluck, M.; and Kosslyn, S. 1984 Pictures 
and Names: Making the Connection. Cognitive 
Psychology 16:243-275. 
Nirenburg, S. and Nirenburg, I. 1988 A Framework for 
Lexical Selection in Natural Language Generation. 
Proceedings of the 12th International Conference on 
Computational Linguistics (2):471-475. 
Nirenburg, S.; Nyberg, E.; and Kenschaft, E. 1987 Inex- 
act Frame Matching for Lexical Selection in Natural 
Language Generation. Unpublished memo, Center for 
Machine Translation, Carnegie-Mellon University. 
Pustejovsky, J. and Nirenburg, S. 1987 Lexical selection 
in the process of natural language generation. In 
Proceedings of the 25th Annual Meeting of the Asso- 
ciation for Computational Linguistics: pages 201-206. 
Reiter, E. 1990 Generating Descriptions that Exploit a 
User's Domain Knowledge. In R. Dale, C. Mellish, 
and M. Zock (Eds.), Current Research in Natural 
Language Generation. Academic Press: London. 
Forthcoming. 
Rosch, E. 1978 Principles of Categorization. In E. 
Rosch and B. Lloyd (Eds.), Cognition and Categori- 
zation. Lawrence Erlbaum: Hillsdale, NJ. 
Zgusta, L. 1971 Manual of Lexicography. Academia 
Press: Prague. 
