Acquisition of Computational-Semantic Lexicons 
from Machine Readable Lexicai Resources 
Jason J. S. Chang and J. N. Chen 
Department of Computer Science 
National Tsing Hua University 
Hsinchu 30043, Talwan, ROC 
Tel: +886 35 731-069 Fax: 723-694 
(jschang,dr818314 }@cs.nthu.edu.tw 
Abstract 
This paper describes a heuristic algorithm capable of automatically assigning a label to 
each of the senses in a machine readable dictionary (MRD) for the purpose of acquiring a com- 
putational-semantic lexicon for treatment of lexical ambiguity. Including these labels in the 
MRD-based lexical database offers several positive effects. The labels can be used as a coarser 
sense division so unnecessarily fine sense distinction can be avoided in word sense disambigu- 
ation (WSD).The algorithm is based primarily on simple word matching between an MRD defi- 
nition sentence and word lists of an LLOCE topic. We also describe an implementation of the 
algorithm for labeling definition sentences in Longman Dictionary of Contemporary English 
(LDOCE). For this purpose the topics and sets of related words in Longman Lexicon of Con- 
temporary English (LLOCE) are used in this work. Quantitative results for a 12-word test set 
are reported. Our discussion entails how the availability of these labels provides the means for 
treating such problems as: acquisition of a lexicon capable of providing broad coverage, sys- 
tematic word sense shifts, lexical underspecification, and acquisition of zero-derivatives. 
1. Introduction 
Treatment of lexical ambiguity such as WSD has been found useful in many NLP applications, includ- 
ing information retrieval (McRoy 1992; Krovetz and Croft 1992) and machine translation (Brown et al. 
1991; Dagan et al. 1991; Dagan and Itai 1994). Recently, various approaches (Dolan 1994; Luk 1995; 
Yarowsky 1992; Dagan et al. 1991 ;Dagan and Itai 1994) to word sense division have been used in WSD 
research. Directly using dictionary senses as the sense division has several advantages. First, sense distinc- 
tion according to a dictionary is readily available from MRDs such as the LDOCE (Longman 1992). Sec- 
ond, indicative words and concepts for each sense are directly available in numbered definitions and 
examples. Lesk (1986) demonstated that dictionary entries can be used to generate signatures of senses for 
WSD. However, using MRD as the knowledge source for sense division and disambiguation encounters 
certain problems. Dolan (1994) observed that sense division in MRD is frequently too free for the purpose 
of WSD. A WSD system based on dictionary senses faces an unnecessary and difficult "forced-choices." 
Most researchers resorted to human intervention to identify and group closely related senses. 
This paper describes a heuristic algorithm capable of automatically assigning a label to each of the 
senses in a machine readable dictionary (MRD) for the purpose of acquiring a computational-semantic lex- 
icon for treatment of lexical ambiguity. Including these labels in the MRD-based lexical database offers 
several positive effects. The labels can be used as a coarser sense division so unnecessarily fine sense dis- 
tinction can be avoided in word sense disambiguation (WSD). The algorithm is based primarily on simple 
word matching between an MRD definition sentence and word lists of an LLOCE (McArthur 1992) topic. 
We begin by giving the details of material used, including the characteristics of definition sentences in 
LDOCE and the organization of words in LLOCE. Next, the algorithm for labeling LDOCE senses is de- 
scribed. An illustrative example demonsu~ates the effectiveness of the algorithm. After describing the al- 
30 
gorithm, the experimental results for a 12-word test set are presented. Our discussion also entails the 
possible implication of the labels to such problems as: acquisition of a lexicon capable of providing broad 
coverage, systematic word sense shifts, lexical underspecification, and acquisition of zero-derivatives at the 
sense level. Moreover, the proposed algorithm is compared with other approaches in available literature. 
Finally, concluding remarks are made. 
2. Identifying the topic of senses 
The labeling of dictionary definition sentences with a coarse sense distinction such as the set labels in 
LLOCE is a special form of the WSD problem. No simple method can solve the general problem of WSD 
for unrestricted text. We will show that this labeling task is made simplex for several reasons. For example, 
consider the definition sentences for the first 5 senses of "bank" in LDOCE: 
1. land along the side of a river, lake, etc. 
2. earth which is heaped up in a l~eld or garden, often making a border or division. 
3.a mass of snow, clouds, mud. etc. 
4.a slope made at Oends in a road or race-track, so that they are safer for cars to go round. 
5.= SANDBANK (a high underwater bank of sand in a river, harbour, etc.). 
First of aLl, only simple words are used in the definitions. Furthermore, the text generation 
schemes are rather regular. The scheme that lexicographers used in generating the definitions above is sim- 
ilar to the DEFINITION scheme described in McKeown (1985). A DEFINITION scheme begins with a 
genus term (that is, conceptual parent or ancestor of the sense), followed by the so-called differentia that 
consists of words: semanficaUy related to the sense to provide specifics about the sense. Those relations be- 
tween the sense and its defining words are reflected in semantic dusters that are termed categorical, func- 
tional, and situational clusters in McRoy (1992). Moreover, those relations have been shown to be very 
effective knowledge sources for WSD (McRoy 1992) and interpretation of noun sequences (Vanderwende 
1994). For instance, land, earth, mass, slope, and sand are the genus terms that are categorically related 
to bank. On the other hand, words in the differentia such as river, lake.field, garden, l~end, road. race-track, 
and harbour are Situationally related to bank through the Location relation. Other keywords such as rOOd, 
and race-tra¢\[~ are related functionally to bank through the PartOfrelation. For the most part, those rela- 
tions exist conveniently among words under the same topic or across cross-referendng topics in LLOCE. 
For instance, most of the above mentioned words are listed under the same topic Ld (Geography) of the in- 
tended label/Ld099, or its cross reference Me (Places). Therefore, these definitions can be disambiguated 
very effectively on the base of similarity between the defining keywords and the words lists in LLOCE. 
2.1. Organizing information in LLOCE 
In this work, the labels used for tagging dictionary definitions are taken from the LLOCE (McArthur 1992). 
Words in LLOCE are organized mainly according to subject matter. Nearly 2,500 sets of related words in 
LLOCE are organized according to 14 subjects and 129 topics (TOP). Cross references (REF) between sets, 
topics, and subjects are also given to show various inter-sense relations not captured within the same topic. 
The cross references in LLOCE are primarily between topics. 
The sets under which the word is listed in LLOCE are considered as the initial candidates for labeling. 
For instance, the Candidates for labeling senses of "bank" are the foUowing 4 set labels: 
Jel04 (banks, exchange, etc.), 
Jel06 (banking and saving), 
Ld099 (fiver banks), and 
Nj295 (bending and leaning) 
The set label Jel04 (as weU as Jel06) is listed under the topic Je (Banking, Wealth, and Investment), while 
31 
Ld099 and Nj295 are listed under Ld (Geography) and Nj (Action and Position) respectively. For instance, 
there is a REF link (in Figure 1) from topic Je to topic De (Belonging and Owning, Getting and Giving). To 
facilitate estimation of similarity between a definition sentence and a topic, we use TOPS to denote the list 
of words under a LLOCE topic S, while REFS denotes the list of words under cross references of S. For 
instance, the label Jel04 (as well as Jel06) is associated with a list of words from its topic (TOPJel04) and 
cross reference (REFJe l 04 = TOPDe): 
TOPJe l04 = TOPJe = {affluent, budget, cut down, deficit, economize, fortune, giro, 
income, keep, luxury, maintenance, needy, pay, windfall, amenity .... } 
REFJeI04 = TOPDe = {bring back, contribution, doff, equip, facility, keep, yield, ... }. 
O Subject 
People Material \ 
Q Topic 
CS> Sets 
O/rganization 
k \ 
\ 
k 
\ 
°oo~ 
Owning ~a~atedal 
-,---_ .... 27z =~ / 
i r/ / 
e- / 
_--¢- 
i~ / / 
_A / ~ ¢- 
p~ 
..... I 
I \ 
~X'h~ -, 
cross-reference 
... b aak... 
Figure 1. Subjects, topics, sets, and cross reference between topics in LLOCE. 
32 
3. The algorithm 
The algorithm is divided into two stages. The preprocessing steps such as part-of-speech tagging, and re- 
moval of stop words are necessary for the algorithm to obtain good results. Various methods for POS tag- 
ging have been proposed in recent years. For simplicity, we adapted the method proposed by Churchl(1988) 
to tag the definition sentence. In the second stage, we select the label which is associated with word lists 
most similar to the definition as the result. We sum up the above descriptions and outline the procedure for 
labeling a dictionary sense. 
Algorithm: I Sense division for a head word h 
Step 1: GiVen a head word h, read its definition, DEFh, from LDOCE. 
Step 2: For each definition D ofDEFh, tag each word in D with POS information.. 
Step 3: Remove all stop words in D to obtain a list of keyword-POS pair, KEYD. i 
Step 4: Lookup LLOCE for headword h to obtain a list of sets SETh that contains h. For each S in 
SETh, compile a set of words TOPS that listed under the topic of S and REFS the set of words 
listed under it cross references. 
Step 5: Compute similarity Sim(D, S) based on Dice Coefficient for all clef'tuitions D ~ DEFh and labels 
S ,SETh. 
Sim (D, S) = 
whereKEYD= the set of POS-keyword pairs in definition D, 
~= the overall relevancy of cross references to a topic, 
wk= 1/the degree of ambiguity of the keyword k, 
In(a, B)= 1 when a • B, 
In(a, B)= 0 when a ~ B. 
Step 6: Assign to D the label S with the maximum value of Sire(D, S) over a threshold. 
Initially, the candidates are limited to the set labels indicated in LLOCE for the head word. If the 
algorithm finds all initial candidates dissimilar, a second run of the algorithm is executed with candidates 
expanded to all topics in LLOCE. 
3.1 An illustrative example 
We illustrate how the algorithm functions using the 5th definition of the word "interest." The preprocess- 
ing stage for definition of word "interest" includes part-of-speech (POS) tagging and stop word removal, 
thereby yielding the following result: 
h = "interest" 
SETinterest = {Fj228, Fb028, Jell2, KaO06 } 
D = "a share in a company business etc. " 
POSD = { a/det, share/n, in/prep, a/det, company/n, business/n, etc./adv } 
KEYD = {share/n, company/n, business/n} 
/KEYDI = 3 
wshare/n 
wcompany/n 
wbusiness/n 
= 1/l{Del05, Hb037, Je114}1 = 1/3 
= 1/1{Cc042, Co292, Jh225}1 = 1/3 
= 1/1{Gh243, Jd138, Jh225}1 = 1/3 
1. In our case. tagging errors have very little negative impact, because words in I.J.,OCE are organized primarily ac- 
cordln~ to topic not part-of-speech. 
33 
TOPFj228 = WFj = 
TOPFb028 = WFb = 
TOPJell2 = WJe = 
TOPKaO06 = WKa = 
{quite/adj, calm/adj .... interest/n, excitement/n, shrill/n .... } 
{likeN, fancy/v .... attraction/n, appeal/n, interest/n .... } 
{lend/v, loan/v .... interest/n, investment/n, share/n .... } 
{entertain/v, amuse/v .... game/n, hobby/n, interest/n .... } 
REFFj228 = WK = WKa ~,Wkb .... WKh 
= { entertain/v, amuse/v, ... game/n, hobby/n, interest/n, .. } 
REFFb028 = WCc = {friend/n, aquaintance/n .... companion/n, company/n .... } 
REFJell2 = WDe = {belong to/v, have/v .... share/n } 
REFKaO06 = WFj = {quite/adj, calm/adj .... interest/n, excitement/n, shrill/n .... } 
I TOPFj228 u REFFj2281= 1693 
I TOPFb028 ~ REFFb0281 = 253 
J TOPJell2 ~ REFJe1121 = 446 
I TOPKaO06 v REFKaO061 =224 
Sim(D, Fj228) 2 
Sire(D, Fb028) 
Sire(D, Jell2) 
Sire(D, KaO06) 
=0 
= 2x0.33x(lxl)/(253+3) = 0.66/256 = 0.00258 
= 2x0.33,,(l+lxl)/(446+3) = 0.66~2/449 = 0.00294 
=0 
The word lists associated with the label Jell 2 is most similar to the key-words of the definition. 
Therefore, the algorithm produces Je112 as the label for "a share in a company business etc." 
3.2 Experiments and Evaluation 
An experiment was carried out using a test set 3 containing 12 polysemous words used in recent WSD ex- 
periments (Yarowsky 1992; Luk 1995). The 12-word test set used in the evaluation represents much more 
difficult cases than average. There are on the average 2.6 definitions in LDOCE for each words as opposed 
to the average 6.4 definitions per words in the test set. Table 1 displays a word by word performance of the 
algorithm. The results show that on the average the algorithm can assign labels to 87% of the senses with 
94% precision. 
4. Discussion 
In this section, we thoroughly analyze the labeling performed by the algorithm and, in particular, look 
into several uses that are made possible by the labels' availability. In addition, those cases when the algo- 
rithm failed can also be analyzed. Analyses result not only illustrate the merits of these labels, but also im- 
ply possible improvement of the algorithm. 
4.1. Broad coverage 
About 50% of the labels are assigned during the second run of the algorithm from the extended candidate 
set. These labels represent gaps in the LLOCE. So, the algorithm can produce much broader coverage than 
the original LLOCE. 
2. For simplicity, the parameter , is set to i. 
3. Only entries relevant to ~e test set m LLOCE are manually emered to ~e computer. We are currently trying to get 
a licence s.t. the full LLOCE entries in order to ccmduct a more complete test. 
34 
Table 1. Performance of the extended algorithm 
headword Alternate sets in No. of deft- Labeling with expanded candidate set 
LLOCE nifions 
inLDOCE La-Incorrect Unknown \[ADolicabili- Precision 
mole 
sentence 
slug 
bass 
bow 
cone 
duty 
galley 
n Ac061 
n Mf159 
nl~.~ 
nLa! 
n Cm256 
n Gf235 
v Gk210 
nAgll3 
n Hh242 
nGd 
nHb 
v blk335 
adj Kb041 
n Kb041 
hE& 
n.Fgt 
n 1-111234 
, n Kb046 
n Mf157 
n Nj295 
v Nj295 
vMb 
vK.h 
nlag 
n Na008 
n Ai134 
n Aj156 
n Jb044 
nRa 
n Fc063 
n Jf160 
nJh 
n Db038 
n Mf157 
n Mf153 
nGd 
11 
2 
Correct 
belling 
1 
1 
1 
1 
1 
1 
1 
0 
1 
1 
1 
1 
2 
1 
1 
0 
0 
Labelling 
0 
0 
0 
0 
0 
0 
0 
0 
0 
0 
0 
0 
0 
0 
0 
0 
0 
0 
1 
0 
0 
0 
1 
0 
0 
0 
0 
0 
0 
0 
0 
0 
Labelling 
0 
0 
0 
0 
0 
0 
0 
0 
0 
0 
1 
0 
1 
0 
0 
0 
0 
0 
0 
0 
0 
0 
0 
1 
0 
0 
0 
0 
0 
0 
0 
0 
Apl: 
ty 
100% 
100% 
80% 
86% 
91% 
100% 
100% 
100% 
100% 
100% 
100% 
100% 
80% 
100% 
100% 
100% 
35 
headword 
interest 
issue 
star 
taste 
Alternate sets in No. of deft- Labeling with expanded candidate set 
LLOCE nitions 
in LDOCE La- Precision 
a Fb028 
n Fj228 
n Ka006 
n Je112 
v Fb025 
v Fj224 
v Ka010 
n.~.£ 
n Aa020 
nGdl80 
n Gf243 
n Nf153 
v Gd174 
n.Ga 
vD¢ 
nlSa 
n_N.~ 
nl',~ 
a Kd082 
n La002 
v Kd079 
nI:lR 
nL~ 
n .N..~ 
n Ia006 
n Nb035 
n .QA.L$1. 
nFI281 
v F1280 
v,n 
v,n 
n.Qb. 
n Fb020 
8 
10 
11 
11 
Correct 
belling 
1 
1 
2 
1 
2 
1 
1 
1 
1 
1 
1 
0 
0 
1 
2 
2 
1 
1 
1 
0 
0 
0 
Incorrect 
Labelling 
4 
2 
1 
0 
0 
0 
0 
0 
0 
0 
0 
0 
0 
0 
0 
2 
0 
0 
0 
0 
0 
0 
0 
0 
0 
0 
0 
Unknown 
Labelling 
0 
0 
0 
1 
0 
0 
0 
0 
0 
1 
0 
0 
1 
0 
0 
0 
0 
0 
0 
1 
1 
1 
0 
0 0 
0 0 
0 0 
0 1 
IlI;'\] 
ApplicabiU- 
ty 
88% 
80% 
73% 
91% 
100% 
75% 
100% 
100% 
II/   -Ills 
Note: Extended labels are underscored. 
4.2. Zero derivation 
Dolan (1994) pointed out that it is helpful to identify zero-derived noun/verb pairs for such tasks as 
normalization of the semantics of expressions that are only superficially different. We have noticed that 
36 
zero derivatives are an important knowledge source for resolving PP-attachment ambiguity. A PP with an 
object involved in a noun/adjective zero-derivation has a strong tendency to attach itself to the preceding 
noun as a modifier. For instance, consider the following example that has an ambiguous PP-attaclmaent 
problem: 
We had a lot of interests in common. 
(= We had a lot of common interests.) 
4.3. Systematic inter-sense relations 
Sanfilippo et al. (1995) contended that strong evidence would suggest that a large part of word sense ambi- 
guity is not arbitrary but follows regular patterns. Moreover, gaps frequently arise in dictionaries and the- 
sauri in specifying this kind of virtual polysemy. Virtual polysemy and recurring inter-sense relations are 
closely related to polymorphic senses that can support coercion in semantic typing under the theory of Gen- 
erative Lexicon of Putstejovsky (1991). 
Our experimental results indicate that the labels in LLOCE make it possible to acquire important inter- 
sense relations, i Many of those relations are reflected in the cross reference information in LLOCE. For 
instance, LLOCE lists the following cross references for the topic of Eb (Food): 
Ac: Animals\]Mammals 
Ad: Birds 
Af: Fish and other (water) creatures 
Ah: Parts of animal 
Ai: Kinds of parts of plants 
Aj: Plant in general 
Jg: Shopkeepers and shops selling food 
most of which are systematic inter-sense relations similar to those described in above-mentioned work. We 
also observed that words involved in such inter-sense relations are frequently underspecified. For instance, 
"chicken" is listed under both topics Eb and topic Ad, while "duck" is listed under Ad but not Eb. By char- 
acterizing of some 200 cross references in LLOCE, most systematic inter-sense relations can be easily iden- 
tiffed among the labeled senses. The labels attached to senses in the MRD, coupled with these inter-sense 
relations, can then support and realize automatic sense shifts advocated in Putstejovsky and Bouillon 
(1994). For instance, the sense of "duck" label with topic Ad can be coerced into an Eb sense when neces- 
sary, with the availability of the lexical rule stipulating a sense shift from Ad and Eb. 
Krovetz (1992) observed that LDOCE indicates sense shifts via direct reference (links indicated 
by a capitalized word with a sense number) and deictic reference (implicit links to the previous sense created 
by this, these, that, those, its, itself, such a, such an). Sense shifts indicated through a deictic reference are 
also present in our 12-word test set. For instance, the first 2 senses of "issue" are 
1. the act of coming out. 
2. an example of this. 
The definition of the 2nd senses indicates an A ctionNoun-CountNoun sense shifts from issue.n.1 to issue.n.2 
through a deictic reference of "this." Since those types of definitions pattern are not considered, the label- 
ing algorithm fails in such cases. Further work must be unde~xaken to cope with direct and deictic referenc- 
es, so that such def'mitions can be appropriately labeled and information on sense shifts can be acquired. 
4.4. Metonymy or Metaphor 
Many definitions indicate metonymical or metaphorical associations between word senses. For instance, 
the 4th and 5th sense of "star" are 
4. apiece of metal in this shape for wearing as a mark of office, rank, honour, etc. 
5. a heavenly body regarded as determining one's fate. 
37 
The 4th and 5th sense are metonymically associated with two "star" senses, star.1 .n.3 (a 5- or more pointed 
figure) and start.l.n.2 (a heavenly body such as a PLANET), respectively. The algorithm often fails in such 
cases for two reasons. First, metonymies are not clearly separated and indicated in LLOCE. Second, the 
genus terms in metonymical senses are often indistinguishable from each other. Further action must be tak- 
en to identify the nature of such relations before this kind of ambiguity can be successfully resolved. The 
presence of phrases "as a ... 03 °' or "regarded as" and drastic change in topic toward the second half of 
the definition may be cues for identifying metonymy and metaphor. 
5. Other approaches 
Sanfilippo and Poznanski (1992) proposed a so-caUed Dictionary Correlation Kit (DCK) in a dialog-based 
environment for correlating word senses across pairs of MRDs, LDOCE and LLOCE. Dolan (1994) de- 
scribed a heuristic approach to forming unlabeled clusters of closely related senses in a MRD. The cluster- 
ing program relies on LDOCE domain code, grammar code, and 25 types of semantic relations exu'acted 
from definitions. Yarowsky (1992) described a WSD method and an implementation based on Roget' s 
Thesaurus and the training material of the 10-rnillion-word Grolier' s Encyclopedia. The author suggested 
that the method can also apply to dictionary definitions. Krovetz (1993) described a simple algorithm based 
on overlap of defining words to identify related senses between morphological variants. The author report- 
ed that the success rate was over 80%. No results were reported for closely related senses within a part-of- 
speech. 
In most of the above-mentioned works, experimental results are reported only for some senses of a 
couple of words. In this study, we have evaluated our method using all senses for 12 words that have been 
studied in WSD literature. This evaluation provides an overall picture for the expected success rate of the 
method, when applied to all word senses in the MRD. Directly comparing methods is often difficult. Nev- 
ertheless, it is evident that in comparison our algorithm is simpler, requires less preprocessing, and does not 
rely on information idiosyncratic to LDOCE. Thus, the algorithm described in this paper can readily apply 
to other MRDs besides LDOCE. Although our algorithm makes use of defining words with various seman- 
tic relations with the sense, explicit computation of those relations is not required. 
6. Conclusions and Future Work 
The meth~ proposed in this work takes advantages of a number of linguistic phenomena: (1) Division of 
senses is primarily along the line of subject and topic. (2) Rather rigid schemes of text generation and pre- 
dictable semantic relations are used to define senses in MRDs such as LDOCE. (3) The implicit links be- 
tween instances of many of these relations are available in a thesaurus such as LLOCE. 
This work also underscores the effectiveness oflexical rules for coarse WSD. Hand-constructed topic- 
based classes of words, coupled with lexical rules as common topic and cross references of topics, prove to 
be highly affecfive both in coverage and precision for WSD, admittedly for sense definitions, a somehow 
restricted type of text. 
Merging senses via labeling has another implication as weU. As discussed in Section 4, the sens- 
es sharing the same label (or cross-referencing labels) are frequently associated through various linguistic 
relations. Making those relations explicit will open the door to flexible treatment of lexicon, semantic typ- 
ing, and semantic under-specification, all of which have received ever-increasing interest. 
In a broader context, this paper promotes the progressive approach to knowledge acquisition for 
NLP as opposed to the "from-scratch" approach. We believe this to be a preferable means to approaching 
a sound and complete knowledge base. 
38 
I 
Acknowledgment 
The authors would like to thank the National Science Council of the ROC for financial support of this re- 
search under Conu'act No. NSC 85-2213-E-007-042. 

References 
1. Brown, P., S.A. Pietra, V.J.D. Pietra, and R. Mercer (1991). "Word Sense Disambiguation using Sta- 
tistical Methods," In Proceedings of the 29th Annual Meeting of the Association for Computational 
Linguistics, pp 264-270. 
2. Church, Ken W. (1988). "A stochastic Parts Program and Noun Phrase Parser for Unrestricted Text." 
In Proceedings of the 2nd Conference on Applied Natural Language Processing (ANLP-88), pp 136- 
143, Austin, Texas, USA. 
3. Dagan, Ido, Alon Itai, and Uldke Schwall (1991). tTwo Languages are More Informative than One,$ Pro- 
ceedings of the 29th Annual Meeting of the Association for Computational Linguistics, pp 130-137. 
4.Dagan, Ido, Alon Itai (1994), "Word Sense Disambiguation Using a Second Language Monolingual Cor- 
pus," Computational Linguistics 20(4), pp 563-596. 
5. Dolan, W.B. (1994). "Word Sense Disambiguation: Clustering Related Senses." In Proceedings of the 
International Conference on Computational Linguistics, pp 712-716. 
6.Kilgarriff, Adam (1993). "Dictionary Word Sense Distinctions: An Enquiry into Their Nature." Com- 
puters and the Humanities, 26, pp 365-387. 
7.Krovetz, R. and Croft (1992). "Lexical Ambiguity and Information Retrieval." ACM Transaction on In- 
formation Systems, pp 115-141. 
8.Krovetz, Robert (1992). "Sense-Linking in a Machine Readable Dictionary." In Proceedings of the 30th 
Annual meeting of the Association for Computational Linguistics, pp 330-332. 
9.Krovetz, Robert (1993). "Viewing Morphology as an Inference Process." In Proceedings of the 16th In- 
ternational ACM SIGIR Conference on Research and Development in Information Retrieval, pp 191-220. 
J. 0. Lesk, Michael E. (1986). "Automatic sense disambiguation using machine readable dictionaries: how 
to tell a pine cone from a ice-cream cone." In Proceedings of the ACM SIGDOC Conference, pp 24- 
26, Toronto, Ontario. 
11. Longman (1992). "Longman English-Chinese Dictionary of Contemporary English." Longman 
Group (Far East) Ltd., Hong Kong. 
12.Luk, Alpha K (1995). "Statistical Sense Disambiguation with Relatively Small Corpora Using Dictio- 
nary Definitions;" In Proceedings of the Annual Meeting of the Association for Computational Linguistics, 
pp 181-188. 
13.McArthur, Tom (1992). "Longman Lexicon of Contemporary English." Longman Group (Far East) 
Ltd., Hong Kong. 
14. McKeown, Katherine R. (1985). "Using Discourse and Focus Constraints to Generate Natural Lan- 
guage Text." Combridge University Press, Cambridge, England. 
15. McRoy, S. (1992). "Using Multiple Knowledge Sources for Word Sense Discrimination." Compu- 
tational Linguistics 18(1), pp 1-30. 
1 6. Putstejovsky, James (1991). "The Generative Lexicon." Computational Linguistics (17)4, pp 409- 
441. 
17.Putstejovsky, James and Pierrette Bouillon (1994). "On the Proper Role of Coercion in Semantic Typ- 
ing." In Proceedings of the International Conference on Computational Linguistics, pp 706-711. 
18. Saniilippo, A., and V. Poznanski. (1992). "The Acquisition of Lexical Knowledge from Combined 
Machine-Readable Dictionary Sources." In Proceedings of the 3rd Conference on Applied Natural 
Language Processing (ANLP-92), pp 80-87, Trento, Italy. 
1 9. Vanderwende (1994). "Interpretation of Noun Sequence." In Proceedings of the International Con- 
ference on Computational Linguistics, pp 454-460. 
20.Yarowsky, David (1992). "Word-Sense Disambiguation Using Statistical Models of Roget' s Catego- 
ries Trained on Large Corpora." In Proceedings of the International Conference on Computational Linguis- 
tics, pp 454.-460. 
