Corpus-Driven Analysis of Multi-Word Terms Including the Word ‘Risk’ in English, French and Lithuanian

8 6 Oksana Smi no a Co pus-D i en Analysis o Mul i-Wo d Te ms Including Sigi a Racke ičienė he Wo d ‘Risk’ in English, F ench and Li huanian Co pus-D i en Analysis o Mul i-Wo d Te ms Including he Wo d ‘Risk’ in English, F ench and Li huanian OKSANA SMIRNOVA Mykolas Rome is Uni e si y, Li huania SIGITA RACKEVIČIENĖ Mykolas Rome is Uni e si y, Li huania KEYWORDS: desc ip i e e minology, co pus d i en analysis, inancial e ms, e m ex ac- ion, e m o ma ion pa e ns 1. INTRODUCTION Compu a ional echnologies ha e opened new possibili ies o linguis ic esea ch: digi al co po a and co pus analysis so wa e acili a ed access o he deepes language s uc u es and ela ions among di e en language uni s and o e eal peculia i ies o hei usage in di e en domains o e di e en pe iods o ime. The e o e, co pus-d i en analysis o na u al language is applied nowadays in a b oad ange o linguis ic a eas: lexico- g aphy, language eaching, de elopmen o machine ansla ion, compila- ion o linguis ic da abases, e c. Te minology esea ch has also become o a la ge ex en co pus-d i en. Digi al co po a allow e minologis s o wo k wi h a big amoun o docu- men s, obse e pa icula ea u es o a specialized language, collec in o - ma ion abou eal usage o e ms and hei e olu ion, cap u e new e ms which could no be in ui i ely el o p edic ed as well as o ca y ou con as i e analysis o da a o se e al languages. Co pus analysis so wa e, au oma ic and semiau oma ic e m ex ac ions ools also con ibu e o e minology s anda disa ion, o example, he equency coun o synonyms can p o ide use ul dis ibu ional e idence indica ing s a is ically p e e ed e ms (Khu shid, Roge s 1992: 36). Thus, digi al co po a enable e mi- nologis s o e ise e minog aphical in o ma ion abou e ms in exis ing da abases and cons an ly upda e hem. 8 7Te minologija | 2018 | 25 Acco ding o M. Te esa Cab é, M. Amo Mon ané and Rogelio Naza , he goal o mode n e minology is o p oduce o mal, seman ic and unc- ional desc ip ions o lexical uni s ha ing e minological alue as well as o explain hei ela ion wi h he es o he uni s o he linguis ic sys em. The objec o e minology esea ch is “ he li ing e minology” ( e min- ology ha na u ally occu s in specialized ex s) and he communica i e aspec o he use o e ms which is bes ins an ia ed in a co pus (Cab é, Mon ané, Naza 2012). The aim, objec and objec i es o he esea ch. The aim o he esea ch is o apply he me hodology o co pus linguis ics o ex ac ion and o mal s uc u e analysis o inancial mul i-wo d e ms including he wo d ‘ isk’ as he head noun in English, F ench and Li huanian. In o de o achie e his aim, he ollowing objec i es we e se : 1) o analyse he p inciples o desc ip i e co pus d i en e minology including he me hods o colloca ional-colliga ional analyses; 2) o compile co po a o he EU legal ac s o inancial domain in h ee languages (English, F ench and Li huanian) and selec he so wa e app op ia e o he co pus-d i en esea ch; 3) o ex ac he mos equen wo ds om he co po a in he in es- iga ed languages and selec he mos equen keywo d (noun) o he u he analysis; 4) o ca y ou colloca ional analysis o he selec ed keywo d in he English co pus and ex ac mul i-wo d e ms including he selec - ed keywo d as he seman ic and syn ac ic head o e ms om he English co pus ma e ial; 5) o es ablish F ench and Li huanian equi alen s o he selec ed English e ms in he pa allel English-F ench-Li huanian co pus; 6) o pe o m o mal s uc u e quan i a i e analysis o he selec ed mul i-wo d e ms and de e mine which modi ica ion pa e ns and syn ac ic s uc u es o he e ms a e p edominan in he in es iga - ed languages. Da a and scope o he esea ch. Fo he pu poses o he esea ch, ou co po a o he EU documen s o inancial domain we e compiled: h ee monolingual co po a (English, F ench and Li huanian) and one pa - allel co pus (EN-FR-LT). The sizes o he co po a a e as ollows: EN 802 933 wo ds, FR 940 655 wo ds, LT 639 279 wo ds. In o al, 210 inan- 8 8 Oksana Smi no a Co pus-D i en Analysis o Mul i-Wo d Te ms Including Sigi a Racke ičienė he Wo d ‘Risk’ in English, F ench and Li huanian cial e ms including he wo d ‘ isk’ as he head noun we e ex ac ed om he co po a: 70 English e ms and hei equi alen s in F ench and Li hua- nian. The choice o wo d ‘ isk’ was de e mined by he co pus da a which e ealed ha his wo d was he mos equen in he selec ed EU documen s. 2. ThEORETICAL pRINCIpLES OF ThE RESEARCh 2.1. P esc ip i e and desc ip i e e minology managemen E olu ion o compu a ional echnologies o na u al language esea ch clea ly di e en ia ed he adi ional e minology om he mode n one. Te minologis s’ a i ude owa ds he concep ion o a e m, e m sou ces, s anda disa ion o e ms and o he e minology issues has changed con- side ably. Two di ec ions o e minology esea ch and managemen ha e been dis inguished: p esc ip i e e minology and desc ip i e e minology (Zel- le 2005; Bielinskienė e al. 2015). Ad oca es o p esc ip i e e minology ocus on e minology s anda di- sa ion based on e minology dic iona ies, da abases, lis s o app o ed e - minology and documen s on s anda disa ion p inciples. In p esc ip i e e minology, concep ion o a e m is based on he place o he concep , desc ibed by i , in he concep ual hie a chical sys em o he domain and i s ela ionship wi h o he concep s (Pea son 1998: 10; Ma cinke ičienė 2000: 6). Acco ding o he p inciples o p esc ip i e e minology, one e m should be used o desc ibe one concep as such ecip oci y educes p obabili y o ambigui y, acili a es communica ion, and, simul aneously, de elopmen o concep ual hie a chical sys em o a domain. Desc ip i e e minologis s dis ance hemsel es om he s ic a i ude owa ds e m concep ion, e m unambigui y, de elopmen o hie a chical concep ual sys em and s anda disa ion. In desc ip i e e minology, con- a y o p esc ip i e one, a con ex plays a i al ole in analysis o a e m, and a e m is assumed as a lexical uni dependen on i s con ex . Thei goal is no o o m a e m acco ding o ce ain p inciples, bu o eco d i s usage, a ie y o i s o ms in di e en ex s and desc ibe i s peculia i ies hus o ming a basis o i s s anda disa ion (Bielinskienė e al. 2015: 10–11). De elopmen o echnologies ga e ise o huge low o in o ma ion p esen ed in a ious ypes o ex s, and, simul aneously, cons an ly g ow- ing numbe o e ms used in hem. In his e e -changing eali y, e mi- 8 9Te minologija | 2018 | 25 nologis s a e no longe able o con ol apid de elopmen o e minology as e ms change as e han hey a e p ocessed. The e o e, managemen o e minology equi es mo e lexible app oach based on desc ip ion a he han on p esc ip ion me hodology. Focus has gone om s and- a disa ion o e minology o he analysis o e minology in co po a (Bielinskienė e al. 2015: 11). Desc ip i e e minology led o de elopmen o he Communica i e The- o y o Te minology which se s ocus on he li ing e minology ha is used in specialized discou ses and pu s he emphasis on he communica i e aspec o he use o e ms (Cab é, Mon ané, Naza 2012). Acco ding o his heo y, he main ole o he e ms is o communica e expe know- ledge; hus hey a e concei ed “as a h ee- old polyhed on ha ing a cog- ni i e componen ( he concep ), a linguis ic componen ( he e m) and a communica i e one ( he si ua ion)” (Cab é, Mon ané, Naza 2012: 1). Te minological s udies based on he communica i e heo y a e no only in e es ed in e minology es ablished by s anda ds o ound in o icial da abases, bu also (and pa icula ly) in hose e ms which a e ac ually used in ex s o language o speci ic pu poses. Thus, he communica i e he- o y “no only adop s an in i o app oach, bu is also in e es ed in e ms in i o” (Cab é, Mon ané, Naza 2012: 1–2). Resea ch o he li ing e - minology discloses ha he adi ional no ion o e m uni oci y (one- o-one co espondence be ween a concep and pa icula e ms in di e en lan- guages) is no well- ounded in eal language. Va ia ion (synonymy, ambi- gui y, pe iph ases, edundancy) is cha ac e is ic no only o gene al language uni s, bu also o e minology; hus concep s and e ms ha e o be s udied “in hei dynamic in e play” (Cab é, Mon ané, Naza 2012: 2–3). The ocus on he usage o e ms has made co po a he main wo kspace o mode n e minology analysis. Digi al co po a and co pus analysis so - wa e ha e enhanced e minology wi h ich esou ces and ools which en- able e minologis s o ex ac e ms om a big amoun o documen s, cap u e he newes changes in e minology o a speci ic domain, analyse i s e olu ion as well as es ablishing concep ual in e connec ion among e ms o he same domain. Co po a allow o ge eliable, s a is ically based, p e i- ously o ally una ailable in o ma ion abou e m usage in na u al language. The e o e, co pus-d i en me hodology has become he p e ailing me hod- ology p o iding indispensable oolki o e minology esea ch and manage- men (Zelle 2005; Cab é, Mon ané, Naza 2012; Bielinskienė e al. 2015). 9 0 Oksana Smi no a Co pus-D i en Analysis o Mul i-Wo d Te ms Including Sigi a Racke ičienė he Wo d ‘Risk’ in English, F ench and Li huanian 2.2. Co pus d i en analysis p inciples: colloca ional and colliga ional me hods Co pus d i en e minology ex ac ion and analysis a e pe o med using s a is ical and linguis ic me hods. In he gi en esea ch colloca ional and colliga ional me hods a e applied. Colloca ions e e o syn agma ic a ac ion be ween lexical uni s. Ac- co ding o Tomas Lehecka (2015), “ he concep o colloca ion is based on he no ion ha each wo d in a language p e e s ce ain lexical con ex s o e o he s, i.e. ha any gi en wo d ends o co-occu wi h ce ain wo ds mo e o en han i does wi h o he s” (Lehecka 2015: 2). S a is ical analy- sis o co pus da a is used o measu e he deg ee o a ac ion be ween wo ds, i s esul s enable o de e mine which wo d combina ions appea oge he signi ican ly mo e o en han i would be expec ed by chance gi en he wo ds’ o al equency in he co pus (Lehecka 2015: 2). In his way he mos signi ican colloca ions o he chosen wo ds in he analysed co pus a e es ablished. This me hod is used mos ly o con ex ual seman ic analysis o lexical uni s as i enables o obse e he whole a ie y o co-occu ing wo ds o he in es iga ed wo ds, es ablish he p edominan co-occu ence pa e ns and hus o desc ibe hei meanings based on hei con ex ual en i on- men (A kins, Rundell, Sa o 2003: 340–341). Colloca ional analysis has become an indispensable ool o lexicog aph- e s, i is also ex ensi ely applied in compu a ional linguis ics o he pu poses o machine ansla ion, na u al language p ocessing and o he a eas (Lehecka 2015). Colloca ional me hod is o en used in combina ion wi h colliga ional me hod. The concep o colliga ion e e s o a ac ion o a lexical uni and a g amma ical pa e n and is based on he no ion ha wo ds p e e o be used in ce ain g amma ical pa e ns and a oid o he g amma ical pa e ns (Sinclai 1998; Lehecka 2015). Colliga ional analysis may been ex ended and encompass he ela ionship be ween he lexical uni and he posi ion in a ph ase, clause, sen ence, ex o discou se whe e he lexical uni can be used (Hoey 2005: 49–52). Colliga ional analysis has been ex ensi ely employed in combina ion wi h colloca ional analysis o s udy he meanings o nea -synonyms as co pus linguis ics s udies ha e e ealed ha hey a e o en used in di e en lexical and g amma ical con ex s (Lehecka 2015). Co pus linguis ics in es iga ions 9 1Te minologija | 2018 | 25 ha e shown ha e en di e en senses o polysemous wo ds may ha e di - e en colloca ional and colliga ional pa e ns (As on, Bu na d 1998). Thus colloca ional and colliga ional analyses ha e p o ed ha seman ics and g amma should no be ea ed as independen , bu a he as closely in e connec ed sys ems (Lehecka 2015). p agma ic aspec should also be aken in o conside a ion in he analyses as colloca ional and colliga ional p e e ences o a lexical uni a y signi ican ly be ween di e en domains and di e en ypes o ex s (Bu le 2004: 157; Newman, Rice 2006). 2.3. Applica ion o colloca ional-colliga ional me hod in e minology ex ac ion and analysis Colloca ional-colliga ional me hod may also be used o ex ac ion o mul i-wo d e ms; i is especially app op ia e o ex ac ion o e min ology including p e-chosen keywo ds – wo ds o po en ial e minological ele ance cha ac e is ic o he domain ha he co pus belongs o. This me hodology is based on he assump ion ha complex e ms a e made o exis ing simple e ms (Nakagawa 2001). Co pus analysis ools ex ac co-occu ing wo ds o he p e-chosen key- wo ds and enable o es ablish he dominan colloca ions. A pa o hese colloca ions is noun ph ases which ha e o be selec ed om colloca ion lis s using linguis ic me hods. The selec ed noun ph ases a e used o u he analysis o he co pus da a and ex ac ion o mul i-wo d e ms consis ing o he selec ed noun ph ases and po en ially o addi ional colloca es. In he gi en esea ch, he ex ac ed e ms a e analysed u he using colliga ional analysis p inciples seeking o e eal o mal s uc u e models o e minology in he in es iga ed languages and o con ibu e o con as- i e mul ilingual s udies o e m o ma ion pa e ns (c . Janule ičienė, Racke ičienė 2014; Mockienė 2016). 3. DEVELOpMENT OF CORpORA AND SELECTION OF ThE SOFTWARE FOR ThE ANALySIS Acco ding o M. Te esa Cab é, M. Amo Mon ané and Rogelio Naza , he pu pose o compiling a co pus o documen s o a speci ic domain is h ee old. Fi s ly, e minologis s need o become amilia wi h he ype o language o he domain; secondly, a co pus is needed o pe o m e - minology ex ac ion and conduc s a is ical analyses o he e ms; and 9 2 Oksana Smi no a Co pus-D i en Analysis o Mul i-Wo d Te ms Including Sigi a Racke ičienė he Wo d ‘Risk’ in English, F ench and Li huanian inally, ex s a e used o ob ain complemen a y in o ma ion abou he e ms such as seman ic, syn ac ic o colloca ional clues (Cab é, Mon ané, Naza 2012: 3–4). The compiled co pus should be o su icien size and quali y o be conside ed ep esen a i e o he chosen domain. The e a e no p ecise equi emen s o he size o a co pus; bu i should con ain as many documen s as possible because he bigge i is, he mo e eliable conclu- sions abou e minology usage in he domain can be made (Cab é, Mon- ané, Naza 2012: 4). Taking in o conside a ion all hese aspec s, he co po a o he EU legal ac s in English, F ench and Li huanian we e compiled. The legal ac s we e downloaded om he O icial Jou nal o he Eu opean Union which is a eely iewed online sou ce. 101 legal ac s, ega ding inancial issues o he EU and enac ed in he pe iod 2014–2017, we e collec ed. The docu- men s we e ans o med in o plain ex and aligned o ex ac ion o e ms and hei analysis. Th ee p og ams we e used o compile, align he ex s and ex ac he necessa y da a om hem: An Conc (2014), An PConc (2017) c ea ed and ce i ied by Lau ence An hony and NOVA Tex Aligne (2014). An Conc is a eewa e mul ipla o m oolki o ca ying ou co pus linguis ic esea ch and da a-d i en lea ning, while An PConc is in ended o c ea ing a pa allel co pus o se e al languages. Bo h p og ams enable he esea che o deal wi h a big amoun o da a and ca y ou a comp e- hensi e mul ilingual linguis ic analysis. The ools which a e p o ided o use s by An Conc p og am a e he ol- lowing: Wo d lis , Colloca es, Clus e s, Keywo ds, Conco dance, Conco dance plo , File iew ool. In he gi en esea ch ou ools o An Conc we e applied: • Wo dlis enabled o de e mine he mos equen wo ds in he co po a; • Colloca es enabled o de e mine he dominan colloca es o he wo d ‘ isk’ and measu e he deg ee o hei syn agma ic a ac ion o he wo d ‘ isk’; • Clus e s allowed o disclose he mos dominan mul i-wo d e ms including he wo d ‘ isk’ as he head noun; • Conco dances ga e a wide a ie y o samples illus a ing he usage o he e ms in he ex s. An Conc p og am’s snapsho is p esen ed in Figu e 1. 9 3Te minologija | 2018 | 25 Be o e compiling a pa allel co pus, he downloaded ex s in h ee lan- guages we e aligned manually wi h he help o he p og am NOVA Tex Aligne . A e alignmen o he ex s, a pa allel co pus was buil up using he An pConc p og am (see Figu e 2). Figu e 2. An PConc p og am Figu e 1. An Conc p og am 9 4 Oksana Smi no a Co pus-D i en Analysis o Mul i-Wo d Te ms Including Sigi a Racke ičienė he Wo d ‘Risk’ in English, F ench and Li huanian The An pConc p og am allowed o expose he examples om h ee languages a he same ime: one could click on a chosen sen ence o see he equi alen s in o he languages. The p og amme also enabled o dis- inguish he keywo ds and hei le o igh colloca es by use o di e en colou s. howe e , i was no possible o see he documen om which he example was ex ac ed. Using he pa allel co pus, F ench and Li hua- nian equi alen s o 70 English e ms we e ex ac ed. 4. CORpUS ANALySIS AND TERM ExTRACTION USINg ThE TOOLS OF ThE pROgRAMS An Conc AND An pConc 4.1. Es ablishmen o he mos equen wo ds in he co po a In o de o de e mine he mos equen wo ds in he co pus o he inancial documen s, he ool Wo d lis o he p og am An Conc was used. I p o ided he wo d equency esul s in he English, F ench and Li hua- nian co po a which enabled o compa e he wo d equencies in he in- es iga ed languages and de elop a ilingual lis o TOp 10 mos equen wo ds (see Table 1). Wo d lis also p o ided in o ma ion on how many wo d okens and wo d ypes he e we e in he co po a and ga e he access o he con ex in which he e ms we e used (see Table 1). Table 1. TOP 10 mos equen wo ds in he co po a EN (co pus 802 933 wo d okens, 9333 wo d ypes) FR (co pus 940 655 wo d okens, 12365 wo d ypes) LT (co pus 639 279, 24727 wo d ypes) 1. isk (3252), isks (592) isque (2855), isques (997) izikos (2549), iziką (1063), izika (536), izikai (355), izikas (4) 2. c edi (3183), c edi s (18) c edi (3307), c edi s (222) k edi o (3103), k edi ų (52), k edi u (29), k edi ą (28), k edi as (8), k edi ai (6), k edi ais (4), k edi us (3), k edi- ams (2) 3. ma ke (1642), ma ke s (424) ma ché (1631), ma ches (484) inkos (1355), inkų (310), inka (229), inkoje (226), inką (76), inko- se (75), inkai (51), inkoms (18), inkas (7), inkomis (2) 101Te minologija | 2018 | 25 Table 6. Syn ac ic s uc u es o he pos nominal modi ica ion pa e ns o he e ms Syn a ic s uc u es EN FR LT ‘ isk’ + pp 13 e ms, e.g.: isk o he ins i u ion, isk o inancial sys em, isk o loss 22 e ms, e.g.: isque de c édi , isque de concen a ion, isque de modèle – ‘ isk’ + A – 20 e ms, e.g.: isque in ajou nalie , isque in e ne, isque spéci ique – ‘ isk’ + A + pp – 4 e ms, e.g.: isque spéci ique de co éla ion, isque signi ica i de co éla ion , isque géné ale de co éla ion – ‘ isk’ + pp + A –15 e ms, e.g.: isque de c édi spéci ique isque de c édi quo idien, isque su ma iè e p emiè e – ‘ isk’ + pp + A + + pp – 9 e ms, e.g.: isque de c édi in ajou nalie à 24h, isque de liquidi é in ajou nalie à 24h – Diag am 1. Modi ica ion pa e ns o he e ms 102 Oksana Smi no a Co pus-D i en Analysis o Mul i-Wo d Te ms Including Sigi a Racke ičienė he Wo d ‘Risk’ in English, F ench and Li huanian The esul s show ha p enominal modi ica ion is dominan in he English and Li huanian languages, while pos nominal modi ica ion is cha ac e is ic o he F ench language. No e ms o p enominal modi ica ion pa e n we e ound in F ench, and no e ms o pos nominal modi ica ion pa e ns we e de ec ed in Li huanian. Only English e ms we e o bo h ypes hough he numbe o e ms wi h pos nominal modi ie s is a he low. All in es iga ed e ms include one o se e al modi ie s which a e nouns, adjec i es o p eposi ional ph ases aking di e en posi ions in he e mino- logical uni s. The esul s o he analysis e eal ha he dominan modi- ie s o he English and he Li huanian e ms a e nouns and adjec i es while in he F ench e ms he wo d ‘ isk’ is mos ly modi ied by p eposi- ional ph ases. The e ms including 3 and mo e wo ds may be u he classi ied acco ding o hei modi ica ion le els ( e ms including modi- ie s modi ying he head noun and e ms including modi ie s modi ying o he modi ie s), bu , due o limi ed space, his analysis is no p esen ed in his pape . The summa ised indings in he ables also e eal ha he e ms di e in he numbe o hei cons i uen s. The quan i a i e analysis o he e m leng h was pe o med o es ablish he dominan numbe o e m cons i uen s in he in es iga ed languages; i s esul s a e p esen ed in he Diag am 2. The diag am e eals wo main endencies. Fi s ly, he EU e m de el- ope s espec he main equi emen o language economy (b e i y o e ms): wo-wo d e ms a e p e alen in all h ee languages. Secondly, only a ew English and F ench e ms ha e mo e han 2-3 wo ds while in Diag am 2. Numbe o e m cons i uen s 103Te minologija | 2018 | 25 Li huanian e ms including 4 and mo e wo ds cons i u e a signi ican pa o he selec ed da a (17 o 70). The endency o p e alence o wo-wo d e ms coincides wi h he en- dencies es ablished by o he e minology esea che s. The esea ch on au oma ic ex ac ion and de ini ion o educa ion and science e minology by Agnė Bielinskienė e al. e ealed ha mos Li huanian e ms o his domain a e wo-wo d e ms: hey cons i u ed mo e han wo- hi ds o he e ms selec ed om he e m candida es au oma ically ex ac ed om a specialised co pus (Bielinskienė e al. 2015: 62). The con as i e esea ch on cons i u ional law e minology by Liudmila Mockienė e ealed he same endency in h ee di e en languages. In he in es iga ed English, Russian and Li huanian legal ac s o a cons i u ional na u e, he majo i y o he ex ac ed mul i-wo d e ms consis ed o wo cons i uen s: hey cons i u ed 78.5% o he English mul i-wo d e ms, 62% o he Russian mul i-wo d e ms and 74.5% o he Li huanian mul i-wo d e ms (Mockienė 2016: 43-45). Thus, de elope s o e ms o di e en domains and di e - en languages end o adhe e o he same p inciple o language economy and use - iendliness. 6. CONCLUSIONS The so wa e An Con , used o monolingual co po a analysis, and An pConc, used o pa allel co pus analysis, allowed o pe o m ex ac ion o English, F ench and Li huanian mul i-wo d e ms including he wo d ‘ isk’ as he head noun om he specialized co po a o he EU documen s o inancial domain compiled o he pu poses o he esea ch. The ex- ac ion was pe o med in he ollowing s ages: • ex ac ing he keywo ds o he English, F ench and Li huanian co - po a ( he wo ds o po en ial e minological ele ance cha ac e is ic o he domain ha he co po a ep esen ) and choosing he mos equen keywo d ( he wo d ‘ isk’) o u he analysis; • ex ac ing colloca ions o he wo d ‘ isk’ and noun ph ases includ- ing he wo d ‘ isk’ as he head noun wi h le and/o igh collo- ca es om he English co pus; • selec ing lexical uni s which ha e e minological alue om he lis o he ex ac ed noun ph ases; • es ablishing F ench and Li huanian equi alen s o he selec ed Eng- lish e ms in he pa allel English-F ench-Li huanian co pus. 104 Oksana Smi no a Co pus-D i en Analysis o Mul i-Wo d Te ms Including Sigi a Racke ičienė he Wo d ‘Risk’ in English, F ench and Li huanian The applied me hodology p o ed o be sui able o e ec i e mul i- lingual ex ac ion o e minology which can be used o de elopmen / upda ing o e mbases o scien i ic analysis o e ms. Fo mal s uc u e analysis o he ex ac ed e ms e ealed some majo e m o ma ion endencies in he in es iga ed languages: • p enominal modi ica ion is dominan in he English and Li huanian languages, while pos nominal modi ica ion is cha ac e is ic o he F ench language; • he dominan modi ie s o he English and he Li huanian e ms a e nouns and adjec i es while in he F ench e ms he wo d ‘ isk’ is mos ly modi ied by p eposi ional ph ases; • he p e alen e m ype acco ding o he numbe o cons i uen s is wo-wo d e ms – hey cons i u e he bigges numbe o he e ms in he TOp 70 lis s in English, F ench and Li huanian; ha shows ha he EU e m de elope s espec he main equi emen o lan- guage economy (b e i y o e ms); • only a ew English and F ench e ms ha e mo e han 2-3 wo ds while in he Li huanian TOp 70 lis e ms including 4 and mo e wo ds cons i u e a signi ican pa o he selec ed da a (17 o 70). The indings o he o mal s uc u e analysis disclose e m o ma ion ends in he in es iga ed languages and p o ide e minological in o ma- ion which migh be use ul o e m de elope s and ansla o s. Syn ac ic pa e ns, es ablished in he esea ch, may be used o de elopmen o au oma ic linguis ic me hods o e m ex ac ion wi hou any p e-chosen keywo ds. REFERENCES As on g., Bu na d L. 1998: The BNC Handbook. Explo ing he B i ish Na ional Co pus wi h SARA, Edin- bu gh: Edinbu gh Uni e si y p ess. A kins S., Rundell M., Sa o H. 2003: The con ibu ion o F ameNe o p ac ical lexicog aphy. – In e na ion al Jou nal o Lexicog aphy 16(3), 333–357. Bielinskienė A., Boizou L, G igony ė G., Ko ale skai ė J., Rimku ė E., U ka A. 2015: Lie u ių kalbos e mi- nų au oma inis a pažinimas i apib ėžimas. Monog a ija, Kaunas: Vy au o Didžiojo uni e si e as. Bu le C. S. 2004: Co pus s udies and unc ional linguis ic heo ies. – Func ions o Language 11(2), 147–186. Cab é M. T., Mon ané M. A., Naza R. 2012: Co pus-based Te minology P ocessing. TKE 2012 Tu o ial. A ailable a h p:// e minus.iula.up .edu/ ke2012/. hoey M. 2005: Lexical P iming: A New Theo y o Wo ds and Language, London: Rou ledge. Janule ičienė V., Racke ičienė S. 2014: Fo ma ion o c iminal law e ms in English, Li huanian and No we- gian. – LSP jou nal – Language o special pu poses, p o essional communica ion, knowledge managemen and cogni ion 15(1), 4–20. 105Te minologija | 2018 | 25 Lehecka T. 2015: Colloca ion and colliga ion. – Handbook o P agma ics Online. J.-O. Ös man, & J. Ve schu- e en (Eds.), Ams e dam: John Benjamins Publishing Company, 1–23. Khu shid A., Roge s M. 1992: Te minology managemen : a co pus-based app oach. – T ansla ing and he Compu e 14, 33–44. Ma cinke ičienė R. 2000: Te minog a ija i eks ynas. – Te minologija 6, 5–23. Mockienė L. 2016: Fo ma ion o e minology o cons i u ional law in English, Li huanian and Russian. Doc o al Thesis, Vilnius: Mykolas Rome is Uni e si y. Nakagawa H. 2001: Expe imen al e alua ion o anking and selec ion me hods in e m ex ac ion. – Recen Ad ances in Compu a ional Te minology. D. Bou igaul , Ch. Jacquenim and M.-C. L’homme (Eds.), Ams- e dam/Philadelphia: John Benjamins Publishing Company, 303–325. Newman J., Rice S. 2006: T ansi i i y schemas o English EAT and DRINK in he BNC. – Co po a in Cog- ni i e Linguis ics: Co pus-Based App oaches o Syn ax and Lexis. S.T. g ies and A. S e anowi sch (Eds.), Be lin/New Yo k: Mou on de G uy e , 225–260. pea son J. 1998: Te ms in Con ex , Ams e dam/philadelphia: John Benjamins publishing Company. Sinclai J. 1998: The lexical i em. – Con as i e Lexical Seman ics. E. Weigand (Ed.), Ams e dam: John Benjamins Publishing Company, 1–24. Zelle I. 2005: Au oma inis e minų a pažinimas i apdo ojimas. Dak a o dise acija, Kaunas: Vy au o Didžiojo uni e si e as, Vilnius: Lie u ių kalbos ins i u as. SOuRCES O icial Jou nal o Eu opean Union: h ps://eu -lex.eu opa.eu/oj/di ec -access.h ml COMPuTER SOTFWARE uSED FOR ThE ANALYSiS An hony L. 2014: An Conc (Ve sion 3.4.4w) [Compu e So wa e]. Tokyo, Japan: Waseda Uni e si y. A ail- able a h p://www.an lab.sci.waseda.ac.jp An hony L. 2017: An pConc (Ve sion 1.2.0) [Compu e So wa e]. Tokyo, Japan: Waseda Uni e si y. A ail- able a h p://www.an lab.sci.waseda.ac.jp Supe no a-so . NOVA Tex Aligne . A ailable a h ps://no a- ex -aligne .so 112.com/. ANgLų, PRANCūzų iR LiETu ių KALbų DAugiAžODžių TERMiNų Su žODžiu izika ANALizė TEKSTYNų LiNg iSTiKOS METODAiS S aipsnyje p is a omi desk ip y iosios e minologijos y imo p incipai bei empi inis daugiažodžių e minų su žodžiu izika y imas, ku io ikslas – aikan eks ynų ling is- ikos me odus, su ink i e minus iš ES inansų s i ies dokumen ų eks ynų i a lik i jų o maliosios sanda os analizę. Ty imo ikslams bu o sukaup i ke u i eks ynai: inansų s i ies dokumen ų anglų kal- ba (802 933 žodžiai), p ancūzų kalba (940 655 žodžiai) i lie u ių kalba (639 279 žo- džiai) bei lygiag e usis anglų–p ancūzų–lie u ių kalbų eks ynas. Iš eks ynų su ink a 210 e minų, ku iuose žodis izika eina pag indiniu dėmeniu: 70 angliškų e minų i po iek pa jų a i ikmenų p ancūzų i lie u ių kalbomis. Žodžio izika pasi inkimą lė- mė ai, kad šis žodis bu o dažniausias isų ijų kalbų eks ynuose. Te minų a pažinimui i su inkimui bu o naudojamos d i kompiu e inės p og a- mos – An Conc i An PConc. Di b a okiais e apais: •dažniausių žodžių, galinčių bū i e minų b anduoliu, angliškame, p ancūziškame i lie u iškame eks ynuose nus a ymas i ieno iš jų (žodžio izika) a inkimas olesnei analizei; 106 Oksana Smi no a Co pus-D i en Analysis o Mul i-Wo d Te ms Including Sigi a Racke ičienė he Wo d ‘Risk’ in English, F ench and Li huanian •žodžio izika kolokacijų i daik a a dinių junginių su pag indiniu dėmeniu izika i jo kai iaisiais bei dešiniaisiais koloka ais nus a ymas angliškame eks yne; •daik a a dinių junginių, laiky inų daugiažodžiais e minais, a inkimas; •a ink ų angliškų e minų p ancūziškų i lie u iškų a i ikmenų nus a ymas. P i aiky a me odologija leido ezul a y iai su ink i daugiažodžius e minus iš dau- giakalbių eks ynų. Tai duoda pag indą eig i, kad ji gali bū i aikoma e minų kaupi- mui bei y imams. Su ink ų e minų o maliosios sanda os analizė a skleidė kele ą s a bių e minų da- ybos endencijų i iamose kalbose: • y aujan is e minų ipas pagal dėmenų skaičių isose ijose i iamose kalbose y a d ižodžiai e minai; ai odo, kad ES e minų kū ėjai laikosi kalbos ekonomi- jos p incipo i s engiasi ku i kuo umpesnius daugiažodžius e minus; • ik kele as angliškų i p ancūziškų e minų u i daugiau kaip 2–3 dėmenis; uo a pu lie u iški e minai, susidedan ys iš 4 i daugiau dėmenų, suda o be eik ke i adalį su ink ų e minų; •anglų i lie u ių kalbų e minų da ybos modeliuose y auja p epozicinė i pos - pozicinė modi ikacija, o p ancūzų kalbos – pos pozicinė modi ikacija; •daugumos anglų i lie u ių kalbų e minų p iklausomieji dėmenys y a daik a a - džiai i būd a džiai, o p ancūzų kalboje – p ielinksninės kons ukcijos. Fo maliosios sanda os analizės ezul a ai su eikia in o macijos, ku i gali bū i nau- dinga e minų kū ėjams i e ėjams. Ty imo me u nus a y i sin aksinių s uk ū ų mo- deliai gali bū i aikomi, ku ian kompiu e inius ling is inius me odus au oma iniam e minų a pažinimui be iš anks o pasi ink ų ak inių žodžių. gau a 2018-05-21 Oksana Smi no a Mykolo Rome io uni e si e as A ei ies g. 20, LT-08303 Vilnius E. paš as osmi no a@m uni.eu Sigi a Racke ičienė Mykolo Rome io uni e si e as A ei ies g. 20, LT-08303 Vilnius E. paš as sigi a. acke iciene@m uni.eu