Te minologija | 2010 | 17 2 1
The ole o cogni i e modelling in
gene al and ha o ames in pa icula
in e minology heo y and p ac ice
WIlly MaR IN
Lexicology/Te minology esea chG oup
V ijeUni e si ei ams e dam
1. INTRODucTION AND SuRVEy
In his pape cogni i emodellingwill be unde s ood as he sys ema ic
ep esen a ion o knowledge wi hin a ce ain subjec o knowledge ield.
The e m model will be aken he e as a backg ound agains which, o a
sys em by means o which, knowledge can be o ganised. A TD (Te mi-
nological Da abase) will, among o he s, be ega ded as he desc ip ion o
a knowledge ield by means o a e minology. A e minological dic iona y,
as opposed o a e minological da abase, will be conside ed as seconda y
o he la e , in o he wo ds, dic iona ies will be ega ded as de i a ions
( on -ends) o unde lying da abases.
The pape i sel will consis o h ee pa s. Fi s o all, di e en le els
o cogni i e modelling will be dis inguished and illus a ed. Secondly, a
ame-based app oach o one o he le els, iz. ha o e ms/concep s,
will be discussed. Thi dly and inally, he app oach ad oca ed will be con-
on ed wi h a se o impo an e minological issues in o de o si ua e
and e alua e i .
2. LEVELS OF cOGNITIVE MODELLING
In o de o unc ion well (so ha knowledge can be [easily] acqui ed
and [p ope ly] used), I will a gue ha a TD needs o be o ganised a , a
leas , h ee le els, iz.
• ha o he domain,
• ha o he da abase and
• ha o he e ms/concep s.
In wha ollows he e o e I will deal wi h
• domain modelling (modelling a he highe le el, he so-called
mac os uc u al le el);
2 2 W. Ma in | The oleo cogni i emodellingingene aland ha o  ames...
• da a modelling (modelling a he in e media e le el, ha o he en i-
ies and ela ions in a da abase, he so-called medios uc u al le el);
• e m/concep modelling (modelling a he lowes le el, ha o he
e ms and he concep s, he so-called mic os uc u al le el).
In o de o make clea wha is mean I will illus a e he espec i e
le els one by one in he nex sec ions.
2.1. Domain modelling
I a e minological da abase is mean o deal wi h knowledge and wi h
i s managemen , hen i has o p o ide o a model ep esen ing he way
how he ‘(sub-)wo ld’, he domain, is o ganised/s uc u ed. So, o in-
s ance, in medicineone basic concep , iz. ha o disease(nosologyin
igu e 1 below) s uc u es he whole ield. I is he cen al o ganising
p inciple wi hin his ‘wo ld’. I one alks abou body-pa s he e, i is be-
cause o he ac ha hey a e/can be a ec ed; i one alks abou o ganisms
he same applies; he apeu ic p ocedu es only make sense when hey e-
e o diseases and so a e symp oms ( indings), causes (e iology) e c.
E e y hing in he medicinal wo ld is linked di ec ly o indi ec ly ( ia
o he concep s) o he cen al concep disease.The amoun and g anula -
i y o he in o ma ion gi en abou o he concep s is de ined by his di ec
o indi ec ela ionship. Domain modelling, he e o e, is acondi iosine
quanonwi hou which i is impossible o cons uc a e minological da-
abase. In igu e 1 a simpli ied schema ic ep esen a ion is gi en o such
domain modelling o medicine[ o mo e de ails see Ma in e.a. 1991].
As one can obse e, all main ca ego ies a e cen ipe ally ela ed o
disease (nosology): ana omy (MEMF) and o ganisms by he ‘a ec ’ ela-
ion, e iology by he ‘caused by’ ela ion, indings by he ‘symp oms’ ela-
ion, he apeu ic p ocedu es by he ‘ ea ’ ela ion e c. Typically hen he
medicinal domain is a domain which is well delinea ed and cen alised: a
domain wi h one cen al/co e ca ego y o which all o he ca ego ies a e
ela ed. I goes wi hou saying ha aking he same objec s and pu ing
hem in o a di e en domain (‘d ugs’ in medicine e sus he same ca e-
go y in pha macy)al e s he s uc u e o he ield and he amoun and
cha ac e o he knowledge ha should be exp essed. Tha is, among
o he s, one o he easons why domain modelling is c ucial when s a ing
wi h he cons uc ion o a TD and dealing wi h knowledge ep esen a ion.
Te minologija | 2010 | 17 2 3
Figu e 1: Example o a ‘delimi ed’ cen alised domain (medicine)
No all domains show he same kind o s uc u e hough. So, o in-
s ance, in igu e 2 an example is gi en o he domain o educa ional
sys ems,which has an embedded o onion-like s uc u e. The e one can
a gue ha all he unics oge he o m he whole (onion) and ha ( he e-
o e) one simply canno es ic onesel o, o ins ance, he inne mos
unic, bu will ha e o selec om all unics (ci cles) i one wan s o come
o g ips wi h he domain o educa ional sys ems as a whole.
As he p eceding examples show, subjec ields/domains a e no always
o de ed hie a chically (acco ding o is-a o pa -o ela ions) as one may
expec a i s sigh , because o biological models wi h hei s ic axo-
nomic o de . Mo eo e , some imes domains a e a he uzzy. Whe eas,
o ins ance, he ea ed domain o medicineis a he well delinea ed, ha
o businessis much less so, such as igu e 3 illus a es.
Al hough one can obse e ha he domain businessimplies he in e ac-
ion be ween a company and bo h i s ex e nal pa ne s and i s in e nal
pa s (P = p oduc ion sec ion, F = inancial sec ion, S = selling sec ion,
A = adminis a i e sec ion) and al hough in bo h in e ac ions selling
2 4 W. Ma in | The oleo cogni i emodellingingene aland ha o  ames...
Figu e 2: Example o an onion-like o ganised domain (educa ional sys ems)
Figu e 3: Example o a di use cen alised domain (business)
(business) and (indus ial) pa ne s/pa s come i s , ye he domain as
such emains di use, al hough i is cen ally o ganised.
Wha e e he domain, ha ing a good insigh in o he supe - o mac o-
s uc u e o he ield is a necessa y condi ion bo h o be e delinea e he
subjec - ield i sel and o o ganise/ ep esen knowledge wi hin ha ield
by means o a TD.
Te minologija | 2010 | 17 2 5
2.2. Da a modelling
unde da a modelling I he e unde s and he modelling o da a as in a
da abase, implying
• he de ini ion o he en i ies in he model and hei ela ionships
(e.g. e ms, concep s, colloca ions and he ela ions/links ha exis
be ween hem) and
• he de ini ion o he da a ca ego ies o he di e en en i ies (bo h
he a ibu es and he [domains o hei ] alues).
Indeed, in a da a model one does no only ha e o make clea wha 
one wan s o ep esen om a con en s poin -o - iew (see 2.1 abo e: he
gene al amewo k o mac os uc u e o he wo ld/domain o be ep e-
sen ed), bu also howone will do so: by means o which o mal objec s,
en i ies and ela ions.
In a p ojec called DOT (ac onym o Du ch Da abank o e heids e -
minologie: Da abase Go e nmen Te minology; see Maks e.a. 2000 and
Maks e.a. 2001) he sys em needed as en i ies: concep s, e ms, colloca ions
and links in o de o ep esen e ms, he use o e ms as in colloca ions,
he ela ionship be ween e ms such as (nea ) synonymy, (nea ) equi alence
and he like. Figu e 4 can gi e an idea o wha is mean .
As one will obse e, di e en g aphic i ms a e used o dis inguish
be ween:
• concep s (c),
• e ms (T) and
• colloca ions (cOLL).
Fu he mo e, a clea dis inc ion is made (see ho izon al b oken line)
be ween e ms and concep s, implying ha concep en i ies co espond
o seman ic uni s exp essed by one o mo e e ms in one o mo e lan-
guages. Te m en i ies ep esen one e m oge he wi h i s ull linguis ic
desc ip ion including i s usage. colloca ion en i ies do he same o col-
loca ions. The e a e se e al kinds o links also: bo h explici ( he ull lines
in he scheme) and implici links ( he b oken lines). An example o an
explici link (one he e minologis has o explici ly ill ou ) is ha be ween
a concep and a e m, o ha be ween a concep and a concep (wi h
alues such as NEARSyN, hyPER, hyPO, REL(ATED)). Implici links
a e links ha he sys em can de i e au oma ically: because o he ac ha
e ms a e linked o concep s and ha p agma ic alues a e speci ied pe

2 6 W. Ma in | The oleo cogni i emodellingingene aland ha o  ames...
Figu e 4: En i ies, links and ela ions in DOT (based on Maks e.a. 2000, see also Maks e.a. 2001 o mo e in-
o ma ion)
e m, he ela ions be ween e ms (bo h in a- and in e lingual ones) need
no be men ioned explici ly, bu can be ‘calcula ed’, leading o ull syno-
nymy, comple e ansla ion equi alence, es ic ed ansla ion equi alence,
and nea ansla ion equi alence.
The ad an age o keeping he concep ual and he linguis ic ( e mino-
logical) le el apa is, among o he s, ha he desc ip ion o a e m in one
language does no in luence he desc ip ion o i s so-called ansla ion
equi alen in ano he language. In o he wo ds, one can wo k now wi h
unilingual en ies, meaning ha he e ms o one language can be de-
sc ibed independen ly om ha o ano he one and ye can be linked
wi h each o he ia he concep ual le el.
In igu e 5 he di e ence be ween unilingual and mul ilingual en ies
in a mul ilingual da abase is schema ically ep esen ed.
Te minologija | 2010 | 17 2 7
unilingual en ies (en ies wi hin one language) can be linked wi h
o he unilingual en ies (en ies om one o mo e languages) wi hou one
language biasing he desc ip ion o he o he .
In mul ilingual en ies one en y con ains all in o ma ion o all lan-
guages. The p oblem hen is ha di e ences a he concep ual le el a e
blu ed i e ms om di e en languages a e ea ed as ansla ion equi -
alen s wi hou being ully equi alen .
The abo e no only makes clea ha one canno cons uc a da a model
wi hou ha ing any no ion abou he ( e ms occu ing in he) domain one
is dealing wi h, bu also ha one should no abs ac away om he asks
one wan s o ca y ou wi h he da abank (as in he case o DOT: com-
pa ing law sys ems and ansla ing ‘go e nmen al’ ex s).
Da a modelling no only comp ises he de ini ion o en i ies, bu ha
o he da a/in o ma ion ca ego ies ha ‘deco a e’ hese en i ies as well. In
he nex sec ion I will deal wi h one o hese en i ies, iz. concep s.
2.3 Concep modelling
In he p eceding sec ion I ha e al eady poin ed a some o he ad an-
ages o a concep ual app oach. One o he p oblems encoun e ed he e is
how o ep esen concep s ( aken as men al building blocks o o ganise
knowledge wi h). I one accep s ha he (concep ual) meaning o a e m
is, as a ule, ep esen ed by i s de ini ion, hen one could ep esen he
meaning/de ini ion/concep exp essed by he e m using a seman ic ne wo k
Figu e 5: Mul ilingual e sus Unilingual En ies
2 8 W. Ma in | The oleo cogni i emodellingingene aland ha o  ames...
as a model (see, o ins ance, F aas 1998: 433 ss.). In Ma in 1998 seman-
ic ne wo ks a e ep esen ed in he o m o ames and, among o he s, used
as de ini ion models. In he nex pa I will u he elabo a e upon he ole
o ames and on ha o a ame-based app oach o e minology.
3. A FRAME-BASED APPROAch TO cONcEPT MODELLING
F ames a e aken he e in he AI sense o he wo d, ollowing he Min-
skyan adi ion (see, o ins ance, Minsky 1975). In his sense hey a e
s uc u es ep esen ing backg ound, implici , s e eo yped knowledge which
is necessa y in o de o unde s and concep s and meaning. AI ames à
la Minsky ha e a slo - ille o ma . F om his poin -o - iew a ame is
a se o gene al concep ual ca ego ies o ela ions (slo s) ollowed by
speci ica ions ( ille s). In o de o make clea wha is mean , I will u n
o a conc e e example.
Seman ic ames a e ype-bound, meaning ha hey a e bound o ce -
ain concep ypes. concep ypes need o ha e been es ablished in he
domain modelling phase (see sec ion 2.1. abo e). Fo ins ance, in he
domain o go e nmen e minology a ype such as allowancewill occu .
The ame-like ep esen a ion o allowancelooks as ollows (see also
Ma in and heid 2001: 58):
Table 1: F ame o he ype allowance
     allowance
Slo PARAPhRASE OF SLOT
goal wha he allowanceis mean o
sou ce who pays he a.
bene icia y who ecei es he a.
eason why he a.is paid
size wha he amoun o he a.is
ime when he a.is paid
pe iodici y how many imes he a.is paid
way in which o m he a.is gi en
condi ion unde which condi ions he a.is gi en
In he wo ld o socialse ices hen, he concep pensionwill be e-
ga ded as a oken o he ype allowance.
Te minologija | 2010 | 17 2 9
The unde lying ame o his ype will consis o he ollowing slo s/
elemen s:
• bene icia y whoge s heallowance?
• sou ce whogi es hea.?
• goal wha is hea.gi en o ?
• eason/g ound onwhichbasisis hea.gi en?
• size wha is hesizeo  hea.?
• pe iodici y howmany imesis hea.gi en?
• ime whenis hea.gi en?
• way inwhich o m(e.g.moneyo o he  alues)is he
  a. gi en?
• condi ion whicha e hecondi ionsunde which hea.is gi en?
A de ini ion de i ed om his ame could ead:
A pensionis an amoun o money, ixed by law o (insu ance) ag eemen , paid
o someone (a pensionableo his widowo o phans)by someone else (a [ o me ]
employe ,an execu i eo ganisa ion),pe iodically (e.g. e e ymon h) o p o ide
o he cos o li ing, a e one has e i ed ei he because o ha ing eached
he ixed age o e i emen o because o in alidi y, i a con ibu ion has been
paid o du ing he e m o o ice.
O cou se bo h he conc e e o m and con en s o he de ini ion i sel
s ongly depend on he use hey a e mean o . howe e , i one akes
o g an ed ha he concep ual meaning o a e m is, as a ule, ep e-
sen ed by i s de ini ion, hen cogni i e models such as ames can ce -
ainly be o g ea help in sys ema ising de ini ions. In he nex sec ion I
will y o make clea ha cogni i e modelling in gene al and a ame-
based app oach in pa icula , go beyond ha , and ha e an impac no
only on he ealm o de ini ions and concep s, bu in ha o e mino-
logical heo y and p ac ice in gene al as well.
4. DIScuSSION: IMPAcT OF cOGNITIVE MODELLING
ON ThEORy AND PRAcTIcE OF TERMINOLOGy
The basic claim pu o wa d in his pape has been he ollowing:
I e minology has o do wi h he
• acquisi ion,
• ep esen a ion and
• applica ion