Senses of Polysemous Nouns: 
Building a Computational Lexicon of Basic Japanese Nouns 
Wakako KUWAHATA, Minako HASIMOTO 
Information-technology Promotion Agency (IPA) 
3-1-38 Shibakoen, Minato-ku, Tokyo 105, JAPAN 
kuwahata@stc, ipa. go. jp, hasimoto~stc, ipa. go. jp 
Fumihiro AOYAMA 
Obirin University 
3758 Tokiwa-ch5 Machida-shi, Tokyo 194-02, JAPAN 
Abstract 
We have constructed the IPA Lexicon 
of Basic Japanese Nouns (IPAL-BN), 
which has a hierarchical structure based 
on the syntactic and semantic proper- 
ties of no,ms. In our lexicon, each lex- 
ical entry consists of subentries, and 
subentries have semantic property in- 
formation. Among these elements, we 
focus here on the subentry description. 
Conventional Japanese dictionaries only 
enmnerate various usages. But it is also 
important to clarify the semantic rela- 
tions between subentries. Thus we have 
developed a method for specifying the 
kind of relationship between subentries, 
using special cognitiw~ devices such as 
metaphor, metonymy, and synecdoche. 
After a brief review of the structure of 
our lexicon, we discuss how the nlethod 
can be applied to the lexical descrip- 
tion. 
1 Introduction 
Tile Information-technology Promotion Agency 
(IPA) i has compiled the IPA Lexicon of 
the Japanese Language for Computers, Basic 
Japanese Verbs (IPAL-BV) (1987) and Basic 
Japanese Adjectives (IPAL-BA) (1990). The 
IPAL-BV contains 861 verbs and the IPAL-BA 
contains 136 adjectives a~s lexical entries. These 
lexicons are available for public use and have been 
widely used in various mfiversity and research in- 
stitute projects that have yielded encouraging re- 
sults. We started work on the IPAL-BN project 
in 1990. hi May 1996, we released tlie third edi- 
tion of tile IPAL-BN, with 1,081 nouns as lexical 
entries, for the public on networks with FTP ser- 
vice. 
The IPAL project is characterized by its linguis- 
tic basis. For example, the hierarchical structure 
i A special juridical body under the jurisdiction 
of the Ministry of International Tradc and Industry, 
Japmt. 
of the IPAL-BN, which consists of lexical entries, 
subentries, and semantic property information, re- 
flects our linguistic considerations concerning the 
syntactic and semantic properties of nouns. An- 
other example of benefits fi'om our linguistically 
inspired approach is the description of the kind of 
relationship between subentries. Such information 
would be useful in various applications, but is not 
yet explicitly provided in existing Japanese dictio- 
naries. In the following sections, we first briefly 
introduce the general structure of the IPAL-BN, 
and then describe our method for specifying the 
kind of relationship between subentries. In the 
concluding renlarks, we also touch on implications 
of the nlethod for tile application systems. 
2 Structure of IPAL-BN 
Figure 1 shows the top-level structure of the 
IPAL-BN. The H'AL-BN consists of 1,081 lexical 
entries. 
IPAL-BN 
Figure 1: Basic Structure of IPAL-BN 
Each lexical entry is conlposed of orthographic 
information, idiomatic information, and suben- 
tries. Tile idiomatic in\]ormation accommodates 
the idiomatic or proverbial uses of the noun that 
have to be treated separately. Let us take an ex- 
ample, Hanako-wa hana-ga takai. This Japanese 
sentence has the idiomatic meaning, "Hanako is 
proud," in addition to the ordinary nleaning, 
"Hanako has a long nose." hi the idionlatic cases, 
the meaning of tile whole sentence cannot be de- 
composed into the meaning of each word. Thus we 
reserve the idiomatic information separately fi'om 
1082 
ordinary meaning sections. 
Then we introduce a hierarchy, subdividing 
each entry that has more than one usage of the 
word. Each usage is (:ailed a subentry. The subdi- 
vision to subentry is based not only on semantic 
but also on syntactic characteristics. This catego- 
rization process can be ilhtstrated with an exam- 
pie of hankyo 'echo': 
hant:y5 q~('he' 
01. a somM t,hat is reflected off a surface such as tit(: 
wall of a lmilding. 
Bbru-no oto-ga hcya-no natca-de hankyS-suru. 
(The sound of the ball echoes in the room.) 
02. one's (,xpr(:ssion of an opinion about or at.tit.mh: 
toward som('thing. 
Kon, o hanashi-wa hitobito-uo kokoro-ni hwnky5- 
o yobiokosu darS. 
(This story will arouse an echo in every man's 
heart) 
One may note that hanky&O1 has a usage in 
which a noun t)ecomes a verb when followed by 
"-suru", while hanky&02 does not. On the basis 
of tlfis (lifference, we divide this i20,112 into these 
two subentries. 
Figure 2 giw.'s an overview of each subc, ntry. 
A subentry consists of subentry information and 
several pieces of semantic property information. 
The subentry information contains syntaeti(:, se- 
mantic, and morphoh)gical information co,ninon 
to all parts of the subentry (each selnantie prop- 
erty ilfforlnation section). The semantic t)roperty 
information in(hides syntactic and semantic in- 
formation. In the case of hanky5, we use the syn- 
tactic information in the subentry information to 
describe, tim difference in the usages (Kuwahata, 
1995). 
Subentry 
----I Subentry information \] 
s .... tic intormatk,~ ) (b) 
Mo,~,,o,o0,~,,,,fo,~ ) (c) 
ic property information /~ 
Syn ac ic o ma ion .................. : (1:: 
-~emantic prope;y information 1;:: 
,;,aciic il;)o;m~iion 
so,.~..~,.,o..~.o. _), 
(a) Syntactic usages 
(1)) Lexical nleanings 
(c) Compound nominals and allomorphemic words 
(d) Regular collocations 
(e) Sen(antic i)rot)erti(,s, synonyms, an(t ant()nylns 
Figure 2: Structure of Subentry 
In addition, we examine the sul)entries in more 
detail and introdu(:e the concept, of the ast)ects 
of nouns. \[,'or example, 'the letter' in 'I read the 
letter' focuses oi1 tim information in the letter, 
whereas its counterpart in 'I l>urned the letter' fo- 
cuses on the thing (i.e., piece of paper) bearing 
that information. Since we can say 'I burned the 
letter that I had read', the word 'letter' does not 
have two meanings but rather has two aspects. 
Tiros a noun is consi<lere<l here to have various 
aspects depending on the predicates used in the 
sentence containing the noun. These aspects are 
called semantic properties (Aoyama, 1995). In- 
stead of subdividing the lexical entry into multiple 
subentries, we categorized the regular collocations 
in each subentry in terms of semantic properties. 
For exan2ple, let, us take the word ha 'tooth 
/teeth' which has three semantic t)rot)erties. In 
our notation, the semantic properties are labeled 
by three letters in square brackets. The phrases 
ha-o migaku 'l)rush one's teeth' and ha-o nuku 
'pull one's tooth' refer to tooth as a concrete ob- 
ject \[CON\]. And the phrases ha-ga haern 'cut 
teetlf and ha-ga nukeru 'h)se teeth' imply (nat- 
ural) phenomena \[PHE\], while the t)hrases ha-ga 
jobu-da 'tlave sound teetlF and ha-ga guragura- 
suru 'a tooth feels h)ose' single out a condition of 
teeth from their potential conditions \[POT\]. 
3 Relationships of Subentries 
3.1 Problem 
Most, existing Japanese dictionaries merely enu- 
merate various usages. But clarifying the seman- 
tic relations between those usages is important. 
For exan2ple, the noun tamago 'egg/spawn' has 
three senses: 
tanuulo 'egg/spawn' 
01. aIl ol)jec~ covered wit, It a hard shell or a mem- 
brane, produced by a female animal. 
Kingffo-ga tarnago-o urtda. 
(The gohlfish spawm:d.) 
02. a heat's egg (i.e., some kind of food). 
Hana~:o-wa tamago-o ichi pa~:ku katta. 
(tlanako bought 1 dozen eggs.) 
03. at person ~tt the Imginning of his/her carter. 
Hanal~o-wa isha-no tam, a.qo-da. 
(Hanako is a dot:for lit the inaking.) 
A hen's egg (02) ix one type of object that 
ix covered wittl a hard shell (01), and (03) is a 
metaphor with respect to the relation between 
hen's egg (02) and hen. Our problem was to ex- 
plicidy (lescribe these kinds of relations between 
meanings. 
3.2 Approach 
Ill the literat, ure, several attempts \]lave al- 
ready been made to analyze such senlantie 
relationships 2. Yanlanasld (1995), among others, 
1)oints out that apt)re(taring such special cogni- 
tive devices as metaphor and metonymy is the 
2Ulhnann, 1969; Lakoff and Johrlson, 1980; Kuni- 
hire, 1982; Yamanashi, 1995. 
1083 
key to understanding polysemy. We emt)loy this 
method for sI)ecifying the kind of relationship l)e- 
tween subentries. Currently we note three types 
of relations: 
• Metaphor, based on similarities 
• Metonymy, based on various (:ontiguitics 
• Synecdoche, l)ase.d oil the relation between 
m(mtl)er mM (:~0;(:gory 
Metaphor is the similm-ity-based instrument 
for exte,Ming the meaning of words. At the out- 
set met, al)horical expressions are temporarily used 
figures of speech. However, some metaphors conm 
to be fixed and pass into everyday use. Let us take 
some examples: h.on-no mushi 'a worm of books' 
(a person who is crazy about reading), arasof 
no tane 'a seed of argument' (a cause for argu- 
ment), and kotoba-no kabe 'a wall of words' (a 
language barrier). We regard these ext)ressions as 
flfll-fledged usages of dm nouns, and accordingly 
descrihe them in the IPAL-BN, while telnporal 
usages are not considered fl)r description in our 
lexicon. 
Metonynly is an instrument for employing a 
word to refer to something that distinct front, hut, 
is associated in some way with, the original ref- 
erent of the. word. Tyi)ieal examples are Nabe-ga 
oishi, "The dish is nice;" and Ano kyateh&wa ( 
kata-o shiteiru, "The catcher has nice shoulders 
(The catcher has a powerflfl throw)." In the for- 
mer, the reference has shifted from the container 
to the content, and in the latter, the refl~.rence has 
shifted Dora the part of the body to its flmetion. 
Synecdoche is the instrument that take.s the 
name of a category to stand for one of its members 
or taking the. name of one ineml)er to stand for the 
whole category, as shown in the above hen's egg 
example: a hen's egg is one kind of object that is 
covered with a hard shell. 
It is lint)or(ant not to confitse the colmoted re- 
latio,l of synecdoche with the contiguity relation 
\[part\]-\[whole\] of lltetollyttly. Let 11s take all ('x- 
ample of the relation \[part\]-\[whole\] to show the 
difference: 
te carln/hmtd' 
0l. upl)cr limbs of a hmnan (th(: (mtire.ty, consisting 
of ~rIn ~tiI(l hmtd). 
"Ha.na~:o-wa ryShS-no te-o hirogcta." 
(Hmmko st)r(.'~t(1 her arms.) 
02. the (:ml of a person's arm (tit(: part consisting 
only of the hand). 
':Hanako-wa tc-o tataita." 
(Ilanako (:lal)I)('d her ttands.) 
We do not consider the. \[part\]-\[whole\] relation 
hetween the arm (01) and t;he hand (02)to be all 
instance of syne(:doehe. Let us compare this ex- 
ample of te 'arm/hand' with the examph; of tam- 
ago 'egg.' Besides a hen's egg, there are tmmy 
other sorts of eggs, such as a turtle's egg, a pi- 
geon's egg, mid a swallow's egg. By contrast of 
the arln as a whole, only the hand receives the 
de.',,}igiiation te. For e.xamph'., we do not also ex- 
press the elbow as te 'arm/hand.' Hence. we con- 
sider the contiguity relation \[part\]-\[whole\] to be 
one type of metonymy. We distinguish it fi'om the 
member-category relation of synecdoche a. 
a.a Method 
In this section, we will show sonm examples of 
description. 
Some metaphorical expressions need some nlod- 
ifier and others do ttt)t. 
kata 'shoulder' 
01. ftana~:o-wa kata-o sukumcta. 
(Hanako shrugged her shouhh:rs.) 
02. Sono uwagi-'no kata-wa 5l:\[. 
(Thai; jacket has wide shoulders.) 
03. Sono yam, a-no t:ata-ni yamagoya-ga aru. 
(There is a lodge on that mountain shonhtcr.) 
((}1) is a non-metaphorical use. ((}2) and (03) 
are metaphorical uses. One may notice that, (02) 
can he used without y@zku-no 'of clothes', but 
((13) requires yama-no 'mountain-.' We have the 
impression that (02) is conventionalized. 
We. classify metaphorical usages into two types: 
conventionalized and novel. To the first group we 
assign expressions that can be used without be- 
ing modified. Such an expression is indicated by 
"A = > B." The second group is indicated by "A 
• ".> B." These mean that A is original and B 
in metaphorical. We refer to the above relation, 
hence, as: "01 ::> 02. 01 "'> 03." 
It is hard to judge which is metaphor if both 
SellSeS &re (;onvelltiotlal. Consider: 
shi'wa ~wrinkle' 
01. Hana~:o-wa me-no mawari-ni shiwa-ga 
dekite~:ita. 
(Hanako has got wrinkles r(mnd her eyes.) 
02. Hanako-wa shatsu-'no shiwa-o aitvn-de 
nobashita. 
(Hanako ironed out the wrinkles in her shirt.) 
We regard (01) as non-metat)horical and (02) 
as metaphorical firsl; group. 
Various contiguity relations cause metonymi- 
cal extension. We single (),at, the relation \[part\]- 
\[whole\] as one. group, and assign the other conti- 
guity relations to a second group. The first group 
is indicated by "A >> B" or "A << B" and 
the second group is indicated by "a \[X\] -> \[Y\] 
B." "A > > B" means extension h'om \[whole\] to 
\[part\] and "A < < B" lile.ans extension in the op- 
posite dire~cdon. "A \[X\] --> \[Y\] B" me.ans that 
the relation \[X\]-\[Y\] links A to B. These relations 
inchlde \[eontainer\]-\[eontent\], \[inaterial\]-\[t)roduct\], \[,neans\]-\[act\]. 
Here is an examl)le. 
aSato (1992) also dist, inguishes I)(:tw(,(m synec- 
doche and mctonymy. But. Ullmmm (1969) and Ya- 
inam~shi(1995) include synecdoche in m('|;onymy. 
1084 
ashi'h:g/foot' 
01. at liml) of animal, whit:h in(:lud(~s the fi)ot and is 
used to snpl)ort th(: I)o(ly and for walking. 
Ninge'u-niwa ashi-ga nihon aru. 
(lh,mans have two h;gs.) 
02. a part of tim foot. 
Kate-no ashi-wa 6kil 
(Ih: has big fi..(.t.) 
03. the way one walks. 
Hanako-wa ash, idori-ga ka,'ui. 
(\[tanako has a light fl:c|;.) 
0d. a t;ransI)orl; servi(:(:. 
7}Lifit-de ashi-ga 'aba'wareta. 
(P('ol)h: w('r(' tl(:privcd of trm,sl)ort by a ty- 
I)hoon.) 
05. (;he h~'g-likc i)ar(; of a nonlivinl~ t;hing, usc(l to 
SUl)port; it;s t)ody. 
So,to isu-ni'wa a,shi-ga gonhon arm 
(That (:hair has four h:'o"s.) 
Whih, (02) m,d (03) arc ,,,ct, o,lymies with rc- 
t,,, (oi), (o4) wi,;h 
t,(, (03) ,,,. (00. W,, 
(;ious as folh)ws: "01 > > 02. 01 \[part of a body\] 
>\[t!unction\] 03 "'" > 04. 01 '" > 05." 
LeA, us now Lllrn 1;o syn('.(:tto(:ht;. It is in(t\[(:a(;ed 
1)y "A 7) B" or "A <; B." Tim former lll0.;H1S (:h?Lt 
A contains 1{, atid |;h('. latter m('.ans (;hat B con- 
t;ai.s A. As Iliell(;i()nc, d :-tl)ov('. taln.ayo 'cgg/sl)aWH' 
has tim following thrt:c sul)en(;ries: (01) ;m objt:ct 
<:ov('.r('.d widl hard shell or a nt(:ml)ran(:: produc(:d 
1)y a f0.nlalc, animal, (02) almn's (',gg, (i.('.., s()Itl0, 
kind of fl)od), (03) a person a(; dm t)(@nning of 
his/her career. While. (02) is a syt.,.(:do(:hc with 
rt'r#u(I to (01), (03) is a mCtal,h()r with r(;sp(,ct 
l;o (02). Wc l'(:fo.r (;o (;tics('. rcl;tdon.q nuin('a'i(:ally~ 
l,hus: "01 Y) 02 ""> 03 ." 
4: Conclusion 
We h;w(: (:Oml)ih'.d (-,he H'AL Basic Japancsc 
Noun,~ (IPAL-HN), which contains dt,l;ail(,d dc- 
sc.ripl, ions of synta(:dc, s(:manti(:, morl)hologi(:al , 
mM idiomatic infl)rma(,ion about: nouus by Olll- 
l)loying a structure l;hat consists of subcntries and 
,ser,.a'ntic properties. Specifically, explicit dcs(:rit)- 
don scmmlfic relations between su|)enLries is a 
novel al)t)roach in the compilation of ,lapaut:se dic- 
(;i()ilario.s for (:onll)uters. 
Nouns are often (:onsid('.r('.d to b(; siml)h'. But; 
sin(:c, ilOllttS llave various lllC;tllill~S ~-i,11t\[ ilS;i,g('.:-;: itl 
Jal);mcs(: l)ro(:(:ssit~g sys(;(:ms no(; all nouns (:an I)(: 
hmMh'.d in a, sl,raight,-\['orward way. Wc 1)tdiev('. 
(-,ha(; l;he c.ompli(:al;(?d functions of tmuns r(~(luir(: 
pro(:(:ssing sy,qt,(~ms (:hat use nol: only syn(;acti(: 
grmnmar, which is alr(:a(ly cnq)loyc.(t in (:onv('.n-- 
donal nal;ura\[ language t)ro(:(;sshlg systems, but 
also syn(;acti(: and S(:lnanl;i(: ilfforma£ion~ of l;h('. 
kind (:onl,;tinc(l in tim IPAL-HN l('.xicon. 
in (;his l)a,t)('.r, we focus(:(1 on (;h(: r(:la(;ionship 1)e- 
I;wc(:tt sul)enl;ric.s au(l h:f(; other r(,.lat, ionships llll- 
l;ou(:h(:d. AmolW; |,he o(,h(:r r<:ladoltshit)s ;tl?('. i(l- 
ionl~t(,i(: cxi)ressions mM scmml(;i(: t)rOl)(:rl,i(;s, t,'or 
eXmnl)lt: , l;hc idiotnatic t:xprcssion a,s'h,i-ga h, ayai 
'l)t: quick (m onc/s ft'.e\[;' mt:ans 'spoil quickly.' We 
rio not m(:ntion in t, hc II'AL-BN l;hal; this exmn- 
pie is mv.l;al)horic expression, since we include, it in 
idiolnadc informa.tion t)ul;side mtbentry. To t;akc. 
anol;he.r cxanq)le ill the semantic property h'.vel, 
the 1)hrase 'cont, act t:he police' rcf0.rs t,o 't)olicc' as 
'n organization \[()RG\], while in the phrase 'go 1,o 
t;hc t)olicc' it, is a location \[I,OC\]. The wor(1 (lo(,s 
not; have two m(,.anings but ral;h(:r hns two so.man- 
l;i(: l)rOl)('.rties. Wt: also do not d(:scril/(~ tam r(:la- 
l:ionship l)(~tWt;('.tl (;h(;lll a,S Ol1('. kind ()f' l/l(~l;()Ilyllty. 
hl(:luding t;h(:s(', points, we. expect fltrl;her (:onsi(l- 
(:radon is n(,.(,.(lcd to flflly dcscril)c the s(:ns(:s of 
l)olys(:m~)us nouns in l;h(: h:xicon. 
Acknowledgments 
The II'AL-BN is (:lie result of the ('J\['orl;s of all of 
our 1)roj('t:t m(;mt)crs, ;tll(l Wc arc. (\[('.('t)ly grateflll 
to t, hcm. This 1)at)(,.r ()w('.s much 1;o (;lw. t;houghtt:ul 
and h0,11fful suggest;ions of Akira Honda. Takashi 
Suzuki and Nol"ihiro Ogata also gave us helpful 
(:ommcnl;s. Wc also thank Kunio Kashino fl)r his 
(:OllS|;l'llt:l;iv(: C()llllll(~.l/t,S O11 (~'arlic.r drafts of this l)W - 
per. 
\[{,(~ f(HTO llCeS 
Aoyama, F., Sose.i-ni motozuku meish'i k~j'.,ts"u,-uo 
tame-no wak',.g'u.mi, I'roc. of IPAL Sympos'ittm '95, 
l)p. 1-(.L (in .\]al)ancsc), IPA, 1995. 
Hasinloto, M.. Kuwahata, W., Murata, K., Aoyama, 
F. and T(moikc, 'l'., Som.e rc',n, arks on ways to com.- 
Irilc ,lapan,:s~ lc:ricons for coulputcrs, Pro(:. o j th(: 
h~.tev..a.tiou.al Worksh.op or| Sharablc Natural Lan- 
guage lb:.~'ource,s, pp. 1115-122, 1994. 
Kunihiro, i\['., hni'ron-uo hoh6 (in Jat)ancsc ), 
Taishukan-shotcn, 1982. 
Kuwahai;a, W., llaslmoto, M. and Mm:ata, K., 
Coltstructiolt of 1PAL-I~N:IPA Le.xicou, of Ba,,;ic 
Japanl:,s'('. No'u,*t,s. Proc. of NI)I'RS '95 'vol. 1, pp. 
137-142, 1995. 
Lakoff, G. and ,l()hlts()it, M., Metaphors We l,iv¢ lly, 
The Uniw:rsity of Chicago, 198(t. 
Sato, N., lb:torikkuoka*tkaku (ht Japanese), Kdansha 
(Jakltjut;slt l\]unko I~alt, 1992. 
Ulhnmm, S.. Semantics: An \[ntroductio,I to th, c 
St:ic'ncl: of Mcwnin.q, (Japanese translal:ion I)y Y. 
lkcgami, putAished by Taishukan-shotcn. 1969). 
YmnaHashi, M., Nmchi bu.upo'u ton (in J~H)ant:s(:). lli- 
tuzi Syobo, 1995. 
Appendix: How to get IPALs 
The II!A L-HN ;I, ttd OLhl:F IPAL.s art,, avaibd)h~ for 
lmlflic us< Anyone. ilfl;crcsl;ed in using the \[PALs 
is invited 1;o cont,u:t ouc of (;1~('. aut, hors of this 
l)almr, lh:sidenl,s oul;sidr ,J~-q)a.n ina, y ol)(;~titl (;he 
copies of the ttla.ltlla,ls (wril;|;en ill .Igtl)a.tlt'.sc ) \[or 
l,\]te \[l)a La l!rc.c, of ohm'go. 
1085 
