BY DETECTING NON-COMPOSITIONALITY OF MULTIWORD
EXPRESSIONS
EXTRAÇÃO DE COMBINAÇÕES LEXICAIS RESTRITAS PELA
DETEÇÃO DA NÃO COMPOSIONALIDADE DE EXPRESSÕES
PLURIVERBAIS *
Joana Veloso
UNIVERSIDADE DO MINHO, PORTUGAL
[email protected]
Alberto Simões
UNIVERSIDADE DO MINHO, PORTUGAL
[email protected]
Álvaro Iriarte S.
UNIVERSIDADE DO MINHO, PORTUGAL
[email protected]
,Q WKLV DUWLFOH DQ HYDOXDWLRQ RI D PHWKRG IRU H[WUDFWLQJ UHVWULFWHG OH[LFDO FRPELnations from parallel corpora by detecting non-compositionality of multiword
H[SUHVVLRQV LQ WUDQVODWLRQ ZLOO EH SUHVHQWHG 7KLV PHWKRG SUHVXSSRVHV WKDW E\
¿QGLQJVHTXHQFHVRIZRUGVZRUGELJUDPVDUHXVHGZKRVHWUDQVODWLRQGRHVQRW
follow a simple word-to-word conversion of the component words, a collocation
LVSUREDEO\SUHVHQW
Keywords: &ROORFDWLRQV7UDQVODWLRQ3DUDOOHO&RUSRUD
Neste artigo apresentamos uma avaliação sobre um método para extrair combinações lexicais restritas a partir de corpora paralelos, pela deteção da não comSRVLFLRQDOLGDGHGHH[SUHVV}HVSOXULYHUEDLVQDWUDGXomR(VWHPpWRGREDVHLDVHQD
presunção de que, encontrando sequências de palavras (são usados bigramas de
palavras) cuja tradução não siga a tradução palavra por palavra dos seus compoQHQWHVpSURYiYHOHVWDUVHSHUDQWHXPDFRORFDomR
Palavras-chave: &RORFDo}HV7UDGXomR&RUSRUD3DUDOHORV
:HDUHJUDWHIXOWRWKHDQRQ\PRXVUHYLHZHUVIRUWKHLUYDOXDEOHFRPPHQWV
5
Joana Veloso / Alberto Simões / Álvaro Iriarte S.
0. Introduction
&ROORFDWLRQVFDQEHGH¿QHGLQYHU\GLIIHUHQWZD\VDFFRUGLQJWRGLIIHUHQW
authors, and the same sequence of words can be considered or not a colloFDWLRQE\GLIIHUHQWUHVHDUFKHUVHYHQZKHQXVLQJDVLPLODUGH¿QLWLRQ
$JUHDWQXPEHURIOLQJXLVWVGH¿QH³FROORFDWLRQ´LQWHUPVRIIUHTXHQF\
That is the criterion used by many authors within the computational linJXLVWLFV FRPPXQLW\ IRU FRUSXVEDVHG DXWRPDWLF FROORFDWLRQ H[WUDFWLRQ
+RZHYHU RWKHU OLQJXLVWV HVSHFLDOO\ OH[LFRORJLVWV DQG OH[LFRJUDSKHUV
believe that the fact that two lexemes frequently co-occur in context is not
DJRRGHQRXJKUHDVRQWREHFRQVLGHUHGDFROORFDWLRQ0HO¶þuk et al.
VWDWH
©.MHOOPHULOOXVWUHXQHVVDLG¶H[WUDWLRQDXWRPDWLTXHGHFROORFDWLRQVG¶XQ
FRUSXVLQIRUPDWLVpVDQVLQWHUYHQWLRQGpFLVLRQQHOOHG¶XQOH[LFRORJXHFHGLFWLRQQDLUHHVWUHPSOLG¶H[SUHVVLRQVFRPPH0U6PLWKZDVDPHPEHUWKHDELOLWLHVDEDGWKLQJHWFTXLQ¶RQWULHQjYRLUDYHFOHVFROORFDWLRQVª
&RVHULXGHQLHVWKDWDOH[HPHFRPELQDWLRQLVD³OH[LFDOVROLGDULW\´RQO\
EHFDXVH ZRUGV DUH RIWHQ FRPELQHG )RU &RVHULX D OH[LFDO VROLGDULW\ LV
explained by lexical restrictions based on the linguistic content of a lexeme
WKDWGULYHLWWRFRPELQHZLWKRWKHUOH[HPHV
©«ODSUREDELOLGDGHVWDGtVWLFDGHODVFRPELQDFLRQHVQRWLHQHSUiFWLFDPHQWH
nada que ver con las solidaridades y no es prueba de su existencia: cavallo
bianco es, probablemente, más frecuente que cavallo sauroSHURHQHOSULPHU
caso, la probabilidad de la combinación depende de la realidad extralinJXtVWLFD HQ HO VHJXQGR HQ FDPELR HVWi GDGD OLQJtVWLFDPHQWH SRU HO FRQtenido de sauro.ª&RVHULX
There are software tools that allow the automatic extraction of frequent
lexical combinations, however, these tools do not distinguish between collocations, idioms, free phrases, compound names, or multiword technical
WHUPV,WVKRXOGEHQRWHGWKDW³UHVWULFWHGOH[LFDOFRPELQDWLRQV´HVSHFLDOO\
FROORFDWLRQVZLWK³IUHTXHQWFRPELQDWLRQV´RIWZRRUPRUHOH[HPHVPXVW
QRWEHFRQIXVHG
,WLVWUXHWKDWDFROORFDWLRQLVDOVRDIUHTXHQWFRPELQDWLRQRIWZRRUPRUH
OH[HPHVEXWIUHTXHQWOH[LFDOFRPELQDWLRQVDUHQRWDOZD\VFROORFDWLRQV$
lexical combination such as ler um livro is a frequent combination, but it is
a completely free phrase, a combination of words formed according to the
Extraction of restricted lexical combinations by detecting non-compositionality of multiword expressions
5
syntactic rules (the verb ler can be combined with everything capable of
EHLQJUHDG+RZHYHUWKHDGMHFWLYHVLQOH[LFDOFRPELQDWLRQVVXFKDVativLGDGHIHEULOPXGDQoDUDGLFDOYRQWDGHORXFDódioPRUWDODPRUFHJR are
¿[HGDOWKRXJKWKH\KDYHWKHVDPHPHDQLQJLQWHQVHDORW7KHVHOH[LFDO
combinations are frequent because the choice of adjective is not free, that is,
the fact that these two lexemes frequently appear combined is not the cause
EXWWKHFRQVHTXHQFHRIWKHPEHLQJDFROORFDWLRQ$ORQVR5DPRV
+RZHYHUIURPDOH[LFRJUDSKLFDOSRLQWRIYLHZWKHVHIUHTXHQWOH[LFDO
combinations, although they are not collocations in the narrow sense that
ZHXVHVKRXOGKDYHVSHFLDOWUHDWPHQWLQGLFWLRQDULHVDQGOH[LFDOGDWDEDVHV
,Q WKLV UHJDUG &RZLHV GLVWLQFWLRQ EHWZHHQ restricted collocations and
open collocations LVXVHIXO&RZLHet al.
,WZLOOWKHUHIRUHEHDSSURSULDWHLQOH[LFRJUDSKLFDOSUDFWLFHWRUHFRJnize the existence of more or less free/restricted lexical combinations and
to register usual or frequent free combinations in dictionaries (Ettinger,
&RUSDV
As we know, stability and semantic specialization are the main features
RIWKLVW\SHRIQRQIUHHSKUDVHV&OHDUO\V\QWDFWLFFRQVWUDLQWVZLOOEHKLJKHU
IRULGLRPVWKDQIRUFROORFDWLRQV)RUH[DPSOHWKHLGLRPperder a cabeça
has more syntactic constraints than the collocation prestar atenção, which
DOORZVVRPHFKDQJHVVHH$JXLODU$PDW+RZHYHUIURPD
lexicographical point of view, it is impossible to establish general syntactic
UXOHVIRUDOOLGLRPVDQGFROORFDWLRQV
The criteria that allow us to determine if a lexical combination was
OH[LFDOL]HGFDQQRWEHPRUSKRORJLFDORUV\QWDFWLFLQQDWXUH7KLVKDVPRUH
to do with the consensus and the memory of the linguistic community that
uses it:
©/H FULWqUH XOWLPH GH Gp¿QLWLRQ G¶XQH XQLWp OH[LFDOH HVW ELHQ LFL SDU H[FHOOHQFHOHFRQVHQVXVGHODFRPPXQDXWpOLQJXLVWLTXH>«@QRQSDVFRPPHHQ
V\QWD[HRXHQPRUSKRORJLHSDUODUHFRQQDLVVDQFHG¶XQHERQQHIRUPDWLRQPDLV
VXUODEDVHGHODPpPRULVDWLRQª3DLOODUG
Ultimately, when it comes to distinguishing between free phrases and
collocations, in addition to the more or less intuitive perceptions of speakers, the use of a foreign language serves to illustrate that choosing a collocDWLYHIRUDEDVHRIFROORFDWLRQLVQRWIUHH&DOGHUyQ7RPDV]F]\N
7KHUHIRUHZHDUHRIWKHRSLQLRQWKDWWKHXVHRIDIRUHLJQODQJXDJHFDQEHXVHIXOWRH[WUDFWFROORFDWLRQVIURPSDUDOOHOFRUSRUD
5
Joana Veloso / Alberto Simões / Álvaro Iriarte S.
)RUWKHSXUSRVHRIWKLVZRUNRXUGH¿QLWLRQRIFROORFDWLRQLVLIDQ\RI
WKHPHPEHUZRUGVRUWKHFRPSOHWHVHTXHQFHLQKHULWDGLIIHUHQWPHDQLQJRI
LWVWKHLUXVXDOVHQVHZKHQXVHGLQFRQMXQFWLRQZLWKDQRWKHUZRUGWKHQWKLV
VHTXHQFHRIZRUGVLVFRQVLGHUHGWREHDFROORFDWLRQ
2XUDVVXPSWLRQLVWKDWJLYHQDPXOWLZRUGH[SUHVVLRQtA and its translation tB, tA is a collocation if tB includes words which are not direct translations of any of tAZRUGV
To test this assumption, we chose a mid-sized corpus, the European
&HQWUDO%DQNFRUSXVIURPWKH3HU)LGHSURMHFW$UD~MR$OPHLGD6LP}HV
'LDV)RUWKHODQJXDJHVZHFKRVHWKH6SDQLVK3RUWXJXHVHODQJXDJH SDLU7KH PDLQ UHDVRQ IRU WKLV FKRLFH LV WKH SUR[LPLW\ RI WKH ODQJXDJHV DQG WKH ELOLQJXDO WUDQVODWLRQ GLFWLRQDULHV ZH KDG DYDLODEOH :H
expect to make further tests on this hypothesis with other language pairs in
WKHIXWXUHQDPHO\LQFOXGLQJ*HUPDQLFODQJXDJHV
As for the text to be analyzed, we selected Spanish word pairs in
which one of the words is an adjective and the other is a noun (in any
RUGHU)RUWKLVSXUSRVHWKH)UHH/LQJ6LP}HV&DUYDOKR3DGUy
6WDQLORYVN\PRUSKRORJLFDODQDO\]HUZDVXVHG)RUHDFKZRUGSDLU
HDFKSRVVLEOHZRUGWUDQVODWLRQZDVORRNHGXSLQWKHWUDQVODWLRQVHJPHQW
This was possible with the help of a Spanish/Portuguese translation dicWLRQDU\$SHUWLXP&RUEt%HOORWHWDO
,I DQ\ RI WKH SRVVLEOH WUDQVODWLRQV IRU ERWK ZRUGV RFFXU WRJHWKHU LQ
DQ\RUGHULQWKH3RUWXJXHVHSDUWWKDWZRUGSDLULVGLVFDUGHG2QWKHRWKHU
hand, if only one of the words has a translation in the Portuguese part, the
Spanish segment is stored, together with a snippet of the Portuguese transODWLRQ7KHQH[WVHFWLRQH[SODLQVWKLVDOJRULWKPLQPRUHGHWDLO
The word pairs obtained, together with the respective translation, were
PDQXDOO\HYDOXDWHGIRUZKHWKHUWKH\DUHRUQRWUHVWULFWHGFRPELQDWLRQV
The evaluation section will discuss the details on the manual assessment
of the obtained results, explaining the main problems found as well as the
IXWXUHHQKDQFHPHQWVWRWKHSURSRVHGDOJRULWKP)LQDOO\DVHWRIFRQFOXVLRQVLVGUDZQIURPWKHREWDLQHGUHVXOWV
1. Related Work
The task of identifying restricted lexical combinations, as we will state
LQWKLVDUWLFOHLVQRWQHZ,WLVDUHOHYDQWSURFHGXUHIRUGLIIHUHQWWDVNVRQ
1DWXUDO/DQJXDJH3URFHVVLQJ1/3OLNHIRUH[DPSOH0DFKLQH7UDQVODWLRQ07ZKHUHLGLRPDWLFH[SUHVVLRQVFDQQRWEHWUDQVODWHGOLWHUDOO\(YHQ
Extraction of restricted lexical combinations by detecting non-compositionality of multiword expressions
5
FROORFDWLRQV QHHG WR EH WUDQVODWHG ZLWK FDXWLRQ )RU H[DPSOH computer
graphics cannot be translated into Portuguese as JUi¿FRVGHFRPSXWDGRU,
not because it is a wrong translation, but because the term that was coined
in the Portuguese community was FRPSXWDomRJUi¿FDJUDSKLFV¶FRPSXWDWLRQ
The easiest way to detect sequences of words likely to be considered as
DFROORFDWLRQRUDWOHDVWDVDFRPSRXQGWHUPLVWRXVHWKH0XWXDO,QIRUPDWLRQ0,RU3RLQWZLVH0XWXDO,QIRUPDWLRQ30,PHWULFV$VWKLVLVMXVWDQ
association measure, authors use these indicators together with other techQLTXHVOLNHWKHXVDJHRISDWWHUQV*XLQRYDUW6LP}HV+RZHYHU
by themselves, these two measures are not enough for the extraction of colORFDWLRQV$SULRUVWXG\3DYHOSUHVHQWVDYDVWDPRXQWRIPHDVXUHV
WKDWFDQEHXVHGWRGHWHFWFROORFDWLRQV1HYHUWKHOHVVPRVWRIWKHPSHUIRUP
badly by themselves, and as presented below, new approaches have been
XVHG
Probably the bigger challenge is to detect idiomatic expressions,
mostly when they can also have a literal meaning (like break the ice that
FDQ EH FRQVLGHUHG OLWHUDOO\ RU QRW /L 6SRUOHGHU SUHVHQW D VHW
of different properties that can be extracted from texts in order to detect
LI WKHVH H[SUHVVLRQV DUH EHLQJ XVHG OLWHUDOO\ 3URSHUWLHV DUH YHU\ GLYHUVH
from the usage of prepositions before or after the expression, to graphs of
FRKHVLRQEHWZHHQWKHGLIIHUHQWVHQWHQFHFRPSRQHQWV7KHVHSURSHUWLHVDUH
WKHQXVHGLQD6XSSRUW9HFWRU0DFKLQHDOJRULWKP7KHVHVDPHDXWKRUV/L
6SRUOHGHUDOVRZRUNHGRQWKHXVHRI*DXVVLDQ0L[WXUH0RGHOV
IRUWKLVVDPHWDVN7KHLUHYDOXDWLRQSRLQWVWRSUHFLVLRQIRUWKHGHWHFWLRQRIOLWHUDOH[SUHVVLRQVEXWRQO\WRGHWHFWQRQOLWHUDOLGLRPDWLF
H[SUHVVLRQV
0X]Q\=HWWOHPR\HUDOVRXVHFODVVL¿FDWLRQWHFKQLTXHVWRGLVWLQJXLVKEHWZHHQLGLRPDWLFDQGQRQLGLRPDWLFH[SUHVVLRQV)RUWKDWWKH\
trained a binary perceptron based on two types of features: lexical features, like the usage of capital letters, and graph features, using relations
LQIRUPDWLRQREWDLQHGIURP:RUG1HWDQG:LNWLRQDU\7KHSHUFHSWURQZDV
WUDLQLQJRQ:LNWLRQDU\ODEHOHGGDWDDQGXVHGWKHQRQODEHOHGGDWDIRUWHVW
SXUSRVHV7KHUHVXOWVJRXSWRRISUHFLVLRQDQGUHFDOORYHU
The most relevant study found using multilingual information for the
GHWHFWLRQRIFROORFDWLRQH[WUDFWLRQ6HUHWDQ:HKUOLGRHVQRWXVH
translation information, but only a parser able to process text in different
ODQJXDJHV7KHH[WUDFWLRQPHWKRGLWVHOIGRHVQRWWDNHDQ\UHDODGYDQWDJH
RISDUDOOHOFRUSRUD
5
Joana Veloso / Alberto Simões / Álvaro Iriarte S.
2XUDSSURDFKXQGHUVWDQGVWKHWUDQVODWLRQDVDIXQFWLRQWKDWFDQVRPHKRZ DVVRFLDWH ³semantic” WR HDFK ZRUG 7KHUHIRUH LI WKH WUDQVODWLRQ
³VHPDQWLF´LVQRWFRPSRVLWLRQDOZHKDYHDFDQGLGDWHFROORFDWLRQ
2. The Hypothesis
The hypothesis we are testing is: if a sequence with two words, an adjective
and a noun, is translated by two other words, and only one of them is a
translation of the original words found in a translation dictionary, then we
KDYHDFDQGLGDWHFROORFDWLRQ
7KLVFDQEHEHWWHUH[SODLQHGXVLQJPDWKHPDWLFDOV\QWD[/HWXVGH¿QH
the 7 function, that translates Spanish words into Portuguese, and the conFDWHQDWLRQRSHUDWRUDGRWZKLFKMRLQVWZRZRUGV
The translation of two words wa and wb is considered to be compositional if
(1) T(wa wb) = T(wa7wb)
or even, T(wa wb) = T(wbT(wa)
Therefore, we are looking for a pair of words (wa Zb) in which one of
them is an adjective and the other a noun, and whose translation does not
IROORZWKHHTXDWLRQSUHVHQWHGDERYH7KDWLVZHZDQWWR¿QGZa and
wb where
7wa . wb) = T(wawc‫ޔ‬7wa wb) = wc T(wb) with T(wa wc ‫ ޔ‬T(wbwc
The extraction algorithm used is very simple, and its main purpose is
to test the hypothesis that the collocation extraction based on non-translaWLRQ FRPSRVLWLRQ LV SRVVLEOH7KH DOJRULWKP VWDUWV E\ LWHUDWLQJ RYHU HDFK
WUDQVODWLRQ XQLW LQ WKH SDUDOOHO FRUSXV$ WUDQVODWLRQ XQLW LV FRPSRVHG E\
a segment SSP for the Spanish language, and a segment SPT for the PortuJXHVHODQJXDJH7KHQHDFKSRVVLEOHELJUDPIURPWKHVHJPHQW6SP is anaO\]HGXVLQJWKH)UHH/LQJPRUSKRORJLFDODQDO\]HUORRNLQJIRUDVHTXHQFH
LQZKLFKRQHRIWKHZRUGVLVDQRXQDQGWKHRWKHUDQDGMHFWLYH1RWHWKDW
DOWKRXJK)UHH/LQJKDVPRGXOHVWRGRSDUWRIVSHHFKWDJJLQJWKH\ZHUHQRW
XVHG1HYHUWKHOHVVZHDUHDZDUHRIWKHSUREOHPVWKLVDSSURDFKDULVHVDQG
ZHZLOOGLVFXVVWKHPODWHU
:KHQVXFKDSDLURIZRUGVLVIRXQGWKHLUSRVVLEOHWUDQVODWLRQVHWVDUH
FRPSXWHG1RWHWKDWHDFKZRUGFDQKDYHPRUHWKDQRQHWUDQVODWLRQDQG
WKHUHIRUHZHQHHGWRFRQVWUXFWDVHWRIWUDQVODWLRQV7KLVWUDQVODWLRQZDV
Extraction of restricted lexical combinations by detecting non-compositionality of multiword expressions
5
GRQHXVLQJWKH$SHUWLXPWUDQVODWLRQGLFWLRQDU\7KHQWKHVHWUDQVODWLRQVHWV
are searched in the target language segment SPT,IDQ\RIWKHZRUGVIURP
ERWKWUDQVODWLRQVHWVRFFXUWRJHWKHUWKHZRUGSDLULVGLVFDUGHG
2QWKHRWKHUKDQGLIRQHRIWKHZRUGVKDVDWUDQVODWLRQLQWKHWDUJHW
segment, but the other does not, the Spanish word pair is saved for manual
DVVHVVPHQW7RJHWKHUZLWKWKHZRUGSDLUDVHJPHQWRI3RUWXJXHVHZRUGVLQ
WKHYLFLQLW\RIWKHIRXQGWUDQVODWLRQLVVHL]HGDQGDOVRVWRUHG7KLVOLVWZDV
WKHQDVVHVVHGPDQXDOO\
3. Assessments and Evaluation
The assessment was performed manually using online resources as refHUHQFHVXFKDV,$7(,QWHU$FWLYH7HUPLQRORJ\IRU(XURSH-RKQVRQ
0DFSKDLODQGERWKSDSHUDQGRQOLQH6SDQLVK3RUWXJXHVHDQG3RUWXJXHVH6SDQLVK GLFWLRQDULHV$ /LQJXLVWLFV 06F VWXGHQW FODVVL¿HG HDFK
word pair manually into one of the following classes:
•
Error: used for all entries whose Spanish and Portuguese segments
DUHQRWUHODWHGZLWKHDFKRWKHU7KLVKDSSHQVPDLQO\EHFDXVHWKH
DSSOLFDWLRQZDVQRWDEOHWR¿QGWKHVHTXHQFHRIZRUGVWKDWLQFOXGH
the translation of the selected pair of words, or because the original
FRUSXVKDGDOLJQPHQWHUURUV
•
Free combination: the pair of words is correctly translated, but
it is not a restricted lexical combination (accordingly with the criWHULRQ ZH GH¿QHG HDUOLHU7KLV KDSSHQV PRVWO\ ZKHQ D SRVVLEOH
ZRUGWUDQVODWLRQLVQRWLQFOXGHGLQWKHXVHGWUDQVODWLRQGLFWLRQDU\
•
Restricted combination: the pair of words is correctly translated,
DQGLWFRUUHVSRQGVWRDFROORFDWLRQRUDTXDVLSKUDVHPH
! " #$ $#$ $ %$&
ZHWRRNWKHGHFLVLRQWRFRQVLGHULWDVDIUHHFRPELQDWLRQ7KLVPHDQVWKDW
RXUHYDOXDWLRQLVOHVVIDYRUDEOHWRRXUK\SRWKHVLV
)URPKHUHRQZHZLOOGLVFXVVHDFKFODVVSURYLGLQJUHDOH[DPSOHVIRU
HDFKRIWKHP
5'
Joana Veloso / Alberto Simões / Álvaro Iriarte S.
3.1. Errors
The errors found are from very different kinds, such us from alignment
problems, some minor bugs in the algorithm implementation, or the lack of
translations from the translation dictionary:
•
The use of a morphological analyzer instead of a part-of-speech tagger
OHDGVWRVRPHH[DPSOHVZLWKYHUEVPLVFODVVL¿HGDVQRXQVDQGRUDGMHFWLYHV1HYHUWKHOHVVFRQVLGHULQJWKDWRXUK\SRWKHVLVLVWKHH[LVWHQFHRI
D VHTXHQFH ZLWK D QRXQ DQG DQ DGMHFWLYH WKH H[DPSOHV FODVVL¿HG DV
DUHVXOWRIWKLVSUREOHPDUHLUUHOHYDQWIRUSURYLQJLW7DEOH()* shows
VRPHRIWKHVHVLWXDWLRQV
+,-./ 01 23,46./7 89 /3:;,<:=8>7 ?@/;/ , A/;-,. 98;4 ?,7 4=7:,B/>.C
interpreted as a noun
•
DEFGH IJKLD
presente DQH[R¿JXUDum modelo
conjunto presente
Se o conjunto apresentar
LQIRUPH¿JXUD
este relatório consta de o
FHUWL¿FDGRIDOOD
YHUL¿FDomRGHRFHUWL¿FDGRIDOKDU
The algorithm, when searching the set of words in the context of the
found translation broke the segment, losing the interesting part of the
WUDQVODWLRQ7KLVWXUQHGWKHDVVHVVPHQWLPSRVVLEOH7KLVZDVDSUREOHP
inherited from the bad segmentations performed by other tools like
WKH VHJPHQWHU WRNHQL]HU DQG WKH VHQWHQFHDOLJQHU )RU H[DPSOH WKH
DOLJQPHQWIRU³DFWLYLGDGHVSHVTXHUDV´ computed by the algorithm was
³GH¿QLWLYDGHDVDFWLYLGDGHVGH´*LYHQWKHPLVVLQJZRUGPDUNHGE\
WKHDVWHULVNWKLVVHJPHQWFRXOGQRWEHFODVVL¿HGFRUUHFWO\DQGWKHUHIRUHIHOOLQWKHHUURUFODVV-XVWOLNHZLWKWKHFDVHDERYHZHGRQRW
KDYHDQ\GHWDLORQWKHYDOLGLW\RUQRWRIWKHK\SRWKHVLV7DEOHVKRZV
IXUWKHUH[DPSOHVRIWKLVVHJPHQWDWLRQSUREOHP$OOWKHVHFDVHVFDQEH
VDIHO\LJQRUHGIRUWKHK\SRWKHVLVWHVW%HWZHHQSDUHQWKHVHVZHVKRZ
MNO ,QWKHVHWDEOHVWKHOHIWFROXPQLVWKH6SDQLVKH[WUDFWHGSDLUDQGWRWKHULJKWLVWKHVHJPHQW
H[WUDFWHGIURPWKH3RUWXJXHVHVLGH,QLWDOLFVZHJLYHHPSKDVLVWRWKHWUDQVODWLRQRIWKHWHUPV
IURPWKH¿UVWFROXPQ
Extraction of restricted lexical combinations by detecting non-compositionality of multiword expressions
55
WKHPLVVLQJZRUGV7KLVVHHPHGWREHDSUREOHPRQWKHFRUSXVVHJPHQWDWLRQDQGDOLJQPHQWSURFHVV
+,-./ P1 23,46./7 89 :;Q><,:/R 7/S4/>:7
tTUTVW VXYZW
Sector de o tabaco em (cru)
derechos humanos
promoção de os direitos de (o homem)
productos pesqueros
mercado de os produtos de (pesca)
Seguridad Alimentaria
Europeia para a Segurança de (a alimentação)
Seguridad Alimentaria
Europeia para a Segurança de (a alimentar)
• 7KHDOJRULWKPLPSOHPHQWDWLRQLVQRWSUHSDUHGWR¿QGDOORFFXUUHQFHVRI
WKHZRUGWUDQVODWLRQV7KLVPHDQVWKDWLIWZRVLPLODUSDLUVRFFXUOLNH
³cantidad superior”DQG³calidad superior”) the algorithm will use the
¿UVWWUDQVODWLRQSDLUWZLFHDOLJQLQJ³TXDOLGDGHVXSHULRU´ZLWK³cantidad superior”DQGQRWWKHFRUUHFW³calidad superior”7KLVLVLQGHHG
a bug introduced by our implementation, but when it was detected it
was too late to perform a complete new extraction and restart the manXDOHYDOXDWLRQ7KHUHIRUHWKH\ZHUHLJQRUHGIRURXUK\SRWKHVLVWHVW
7DEOH VKRZV VRPH RI WKHVH H[DPSOHV ,Q LWDOLFV RQ WKH ULJKW WKH
DOLJQHGVHJPHQW
+,-./ [1 23,46./7 89 4=7,.=S>/R 7/S4/>:7
•
\H]XQLGDG(FRQyPLFD
que institui a Comunidade Europeia
cantidad superior
em uma TXDQWLGDGHLQIHULRU
tiempo completo
de trabalho a tempo parcial
tercera columna
referidas em a coluna 2
QDYHJDFLyQPDUtWLPD
afectos a a navegação aérea
The translation, sometimes, uses a pronoun to refer to a noun used on
DSUHYLRXVVHQWHQFHZKLOHWKHRULJLQDOVHQWHQFHUHSHDWVWKHQRXQ6HH
7DEOHIRUVRPHH[DPSOHV
5^
Joana Veloso / Alberto Simões / Álvaro Iriarte S.
Table 4: Examples of misalignments resulted from the use of pronouns
Estado membro
legislação de esse Estado
ciertos productos
regime de esses produtos YDORUHVOtPLWH
ou de esses valores ,
último caso
(Pesse caso ,
segundo Estado
legislação de HVVH(VWDGR
• 6RPH WUDQVODWLRQ XQLWV ZHUH QRW UHDOO\ WUDQVODWHG ,Q VRPH FDVHV WKH
Portuguese version included the text in Spanish, and in some other,
LQ (QJOLVK DV VKRZQ LQ7DEOH 6RPH RWKHUV DV VKRZQ LQ7DEOH include typos that, not being in the dictionary, activated our hypothesis
E\PLVWDNH
+,-./ _1 23,46./7 89 :;,>7.,:=8> Q>=:7 ?@/;/ ,: ./,7: 8>/ 89 :@/ 7=R/7 =7
untranslated
medio ambiente
contaminación del medio ambiente
Autoridades nacionales
Lista de las autoridades nacionales
medio ambiente
en el medio ambiente acuático
Vivo Test
,Q9LYR7HVWIRU&KURPRVRPDO
Table 6: Examples of translations with minor typos, mistakenly extracted as
collocations
SUHVHQWHDUWtFXOR
de o presnete artigo
OHJtWLPRWLWXODU seu legítimo titual a ocupar
]RQDJHRJUi¿FD
HVSHFt¿FDVHPXPD]RQDJHRJUi¿UFD
presente Reglamento
2presidente regulamento é
proyectos transnacionales
acompanhamento de os projectos trannacionais
Extraction of restricted lexical combinations by detecting non-compositionality of multiword expressions
5`
3.2 Free Combinations
The main interference with the algorithm, which could make it extract free
combinations, is the lack of translations from the used translation dictioQDU\:KHQDZRUGLVQRWIRXQGLQWKHGLFWLRQDU\HLWKHUDVWKHZRUGQRW
existing in the source language – 7DEOH RU WKH WDUJHW ODQJXDJH GR QRW
include the used translation – 7DEOHWKHDOJRULWKPFRQVLGHUVWKHWUDQVODWLRQWREHLQFRUUHFWDQGWKHUHIRUHLWFDQEHXVHGIRURXUK\SRWKHVLV$VLPilar problem occurred with words not correctly lemmatized, and therefore,
QRWIRXQGLQWKHWUDQVODWLRQGLFWLRQDU\
Table 7: Basic examples where the dictionary lacked a entry for one of the words
Segundo resultado
resultado : segundo resultado :
primer trimestre
em o primeiro trimestre de
presente Directiva
requisitos de a presente directiva
primeros párrafos
os dois primeiros parágrafos podem
presente apêndice
o presente apêndice , os
Table 8: Basic examples where the dictionary lacked one of the synonyms
Texto pertinente
EEE ) ( 7H[WRUHOHYDQWH
siguiente texto
a seguinte redacção :
0HGLGDVYLJHQWHV
0HGLGDVH[LVWHQWHV
última fabricación
a última data de fabrico
&RQRFLPLHQWRVVXFLQWRV
Conhecimentos sumários de as
)LQDOO\ WKHUH LV \HW DQRWKHU SUREOHP UHODWHG ZLWK WKH WH[WXDO GHL[LV,
where there is a reference to a different position in the text that, different
ODQJXDJHVUHIHUWRLQGLIIHUHQWZD\VExamples are cuadro anterior/TXDGUR
acima, fórmula anterior/fórmula acima and criterios anteriores/critérios
acima. As anterior and acima are not direct translations, the algorithm
extracted them as restricted combinations (although they are free combiQDWLRQV
5
Joana Veloso / Alberto Simões / Álvaro Iriarte S.
3.3 Restricted Combinations and Collocations
2WKHUWKDQWKHFRUUHFWUHVWULFWHGFRPELQDWLRQVWKHUHDUHWZRVSHFLDONLQGV
that should be mentioned:
•
There are situations where the pair of words in Spanish has a single
word translation in Portuguese, either because in Portuguese one of the
words is usually omitted, or because there is a word with the complete
PHDQLQJRIWKHWZR6SDQLVKZRUGV7KLVVLWXDWLRQZDVQDPHGreduction
DQGKDSSHQVDIHZWLPHV([DPSOHVDUHVKRZQLQ7DEOH7KHVHZHUH
FRQVLGHUHG UHVWULFWHG FRPELQDWLRQV7KH EHVW H[DPSOHV IURP7DEOH DUHWKH¿UVWDQGWKHODVW,Q3RUWXJXHVHDQGDOWKRXJKWKHUHLVWKHFRQcept of meio ambiente, it is usually used only as ambiente. And in the
case of FLJDUURVSHTXHxRV, Portuguese as a word for that: cigarrilhas
7KHVHVLWXDWLRQVZHUHYDOLGDWHGPDQXDOO\LQ,$7(,QWHU$FWLYH7HUPLQRORJ\IRU(XURSH
+,-./ a1 23,46./7 89 ;/RQ<:=8>71 7=:Q,:=8>7 ?@/;/ :?8 ?8;R7 ,;/ <8;;/<:.C
translated by only one word
•
medio ambiente
protecção de o ambiente ,
DXGLWRUtDPHGLRDPELHQWDOHV
de ecogestão e auditoria (
titular opositor
parte de o titular
trabajo anual
uma unidade de trabalho ,
FLJDUURVSHTXHxRV
cigarrilhas e cigarros ,
There is another situation with nouns (mostly geographic) that were
PLVWDNHQO\ H[WUDFWHG DV VKRZQ LQ 7DEOH 7KHVH ZHUH H[WUDFWHG
EHFDXVH RI WKH ZD\ WKH QRXQV DUH WUDQVODWHG7KLV WDEOH VKRZV WKUHH
FROXPQV 7KH ¿UVW WZR DUH 6SDQLVK DQG 3RUWXJXHVH DQG WKH WKLUG D
GLUHFW3RUWXJXHVHWUDQVODWLRQRIWKH6SDQLVKWHUP
Extraction of restricted lexical combinations by detecting non-compositionality of multiword expressions
5
Table 10: Examples of nouns whose translation was a problem for the algorithm
Sudeste Asiático
Ásia de o Sudeste
Sudoeste Asiático
5HS~EOLFD)HGHUDO
RULJLQiULRVGHD5HS~EOLFD)HGHUDWLYD
República Federal
FRQWLQHQWDOHVSDxRO
Espanha continental
continental espanhol
• *LYHQWKDWZHGHFLGHGWRDQDO\]HELJUDPVWKHUHDUHVLWXDWLRQVZKHUH
WKHELJUDPLVSDUWRIDELJJHUUHVWULFWHGFRPELQDWLRQ7KLVLVXVXDOO\
HDV\WRGHWHFWJLYHQWKHVSHFL¿FDUHDRIWKHXVHGFRUSRUDDQGJLYHQWKDW
the Portuguese segment includes more words than the two existing in
6SDQLVK7DEOHVKRZVH[DPSOHV
+,-./ 001 b,7/7 89 ;/7:;=<:/R <84-=>,:=8>7 ?=:@ 48;/ :@,> :?8 ?8;R7
gestión medioambiental
o sistema de gestão ambiental
producción homogénea
A unidade de produção homogénea
SROtWLFDDJUtFROD
de a política agrícola comum
ejecución forzosa
de PHGLGDVGHH[HFXomRIRUoDGD
residuos radiactivos
Gestão de os resíduos radioactivos
:KHQ WKH DOJRULWKP UHWXUQHG LQWHUHVWLQJ UHVXOWV UHWXUQLQJ UHVWULFWHG
lexical combinations, we were unable to distinguish between collocations
and other types of restricted combinations, like quasi-phrasemes and idioms (this last type was not found, probably given the type of the used
FRUSXV
5
Joana Veloso / Alberto Simões / Álvaro Iriarte S.
Table 12: Examples where the algorithm worked
Disposiciones legales
disposições legislativas
mercado interior
mercado interno
cadena alimentaria
&DGHLD$OLPHQWDU
Derecho interno
direito nacional
derechos humanos
direitos fundamentais
contingentes arancelarios
contingentes pautais
fronteras interiors
fronteiras internas
entidad contratante
entidade adjudicante
GtDVKiELOHV
dias úteis
DxRVQDWXUDOHV
anos civis
SHUVRQDItVLFD
pessoa singular
peso neto
SHVROtTXLGR
DxRVQDWXUDOHV
anos civis
precio neto
SUHoROtTXLGR
SHUVRQDVMXUtGLFDV
pessoas colectivas
amarillo oscuro
amarelo torrado
VHQWHQFLD¿UPH
sentença transitada
petróleo crudo
RXUHVtGXRVGHSHWUyOHREUXWR
atún rojo
atum rabilho
correo normal
correio ordinário
historiales médicos
processos médicos
3.4 Analysis
:KHQDSSOLHGWRWKH(XURSHDQ&HQWUDO%DQNFRUSRUDRXUDSSURDFKH[WUDFWHG
PRUHWKDQSDLUVRIFDQGLGDWHFROORFDWLRQV7KH\ZHUHHYDOXDWHGE\
exhaustion (instead of evaluating a sample of random entries, the evaluator
WDJJHGHDFKRQHRIWKHH[WUDFWHGFDQGLGDWHV7KLVPHDQVWKHHYDOXDWLRQLV
QRWDIIHFWHGE\VDPSOHELDV7KLVWRJHWKHUZLWKWKHIDFWWKDWWKHHYDOXDWRU
gave preference to free combinations over restricted combinations, means
WKDWWKLVHYDOXDWLRQLVWKHEDVHOLQHRIWKHDOJRULWKP
7DEOH SUHVHQWV WKH QXPEHU RI FDVHV IRXQG DQG FODVVL¿HG DFFRUGLQJWRHDFKRIWKHSUHYLRXVO\PHQWLRQHGFODVVHV,IZHLJQRUHWKHFDVHVRI
errors, nouns and reductions, we can note that restricted combinations are
RQHTXDUWHURIWKHWRWDOQXPEHURIIRXQGFRPELQDWLRQV
Extraction of restricted lexical combinations by detecting non-compositionality of multiword expressions
5
Table 13: Summary of cases found for each category.
&DWHJRU\
)UHH&RPELQDWLRQV
5HVWULFWHG&RPELQDWLRQ
Errors
Reductions
Nouns
1XPEHURI2FFXUUHQFHV
Percentage
19
4. Conclusions
7KH¿UVWUHDFWLRQWRWKHUHVXOWVZDVRIGLVFRQWHQWDVDORWRIIUHHFRPELQDWLRQVZHUHIRXQG$VVRRQDVWKHH[DPSOHVZHUHDQDO\]HGZDVUHDOL]HG
¿UVWO\WKHWUDQVODWLRQUHVRXUFHODFNVFRYHUDJHDQGVHFRQGO\WKHDOJRULWKP
XVHGPLVVHVWKHFRUUHFWOHPPDWL]DWLRQIRUVRPHZRUGV7KHVHWZRUHDVRQV
FDQEH¿[HGRUPDGHEHWWHUXVLQJRWKHUDSSURDFKHVRUWRROVIRUWKHOHPmatization, and using other dictionaries or even probabilistic translation
GLFWLRQDULHV 6LP}HV $OPHLGD 6LP}HV$OPHLGD 5DPRV &DUYDOKRWRHQULFKWKHWUDQVODWLRQFRYHUDJH
Nevertheless, most of the situations found are easy to correct, and,
therefore, further experiments should be performed before considering the
PHWKRG LQDGHTXDWH ,Q IDFW ZLOO EH LQWHUHVWLQJ WR VHH KRZ WKLV DSSURDFK
performs in a less noisy corpus, with better dictionaries, and with other
ODQJXDJHV
7KLV GDWD EHLQJ PDQXDOO\ FODVVL¿HG FDQ EH XVHG WR WUDLQ PDFKLQH
OHDUQLQJ DOJRULWKPV )RU VWDWLVWLFDO PDFKLQH WUDQVODWLRQ LW LV SRVVLEOH WR
denote/specify which segments should be reused directly without any
change (when they are idiomatic), and which segments can be generalized,
allowing some of the words to be replaced, and reusing the translation
VWUXFWXUH)RUWKHH[WUDFWLRQRIIXUWKHUFROORFDWLRQVIURPRWKHUFRUSRUDWKLV
data can be used to train a supervised machine learning algorithm, or just
EHXVHGDVDJROGHQVWDQGDUGIRUWKLVNLQGRIV\VWHP
Analyzing the results obtained, the initial starting hypothesis should
EHUHIRUPXODWHG8VLQJWKLVDSSURDFKUHVWULFWHGFRPELQDWLRQVDQGQRWMXVW
DVSHFL¿FW\SHRIUHVWULFWHGFRPELQDWLRQVVXFKDVWKHFDVHRIFROORFDWLRQV
DUHGHWHFWDEOH2IFRXUVHDXWKRUVOLNH0HO¶þXNGH¿QHIRUPDOW\SHV
IRUHDFKRQHRIWKHVHUHVWULFWHGOH[LFDOFRPELQDWLRQV7KHSUREOHPLVWKH
5
Joana Veloso / Alberto Simões / Álvaro Iriarte S.
QRQH[LVWHQFHRIDFOHDUGLVWLQFWLRQEHWZHHQWKHP6RPHOH[LFDOFRPELQDWLRQVZLOOEHFODVVL¿HGGLIIHUHQWO\DFFRUGLQJWRWKHZD\WKHOLQJXLVWGHFRPSRVHVVHPDQWLFDOO\WKHH[SUHVVLRQ
There is another problem with our hypothesis, when a restricted combination coincides in the two languages being analyzed, because they can
EHPLVWDNHQO\FRQVLGHUHGIUHHOH[LFDOFRPELQDWLRQVExamples of this problem are GHVHJXQGDPDQRHPVHJXQGDPmRyGLRPRUWDOyGLRPRUWDO, amor
FLHJRDPRUFHJR:HH[SHFWWKDWWKLVPD\QRWEHDSUREOHPZKHQXVLQJ
parallel corpora including more distinct languages (Portuguese/English,
6SDQLVK(QJOLVKHWF
References
AGUILAR-AMAT,$/DVFRORFDFLRQHVGHQRPEUH\DGMHWLYR8QSDVRKDFLDXQD
WHRULDOp[LFRVHPiQWLFDGHODWUDGXFFLyQ7HVHGHGRXWRUDPHQWR>PLFUR¿FKDV@
%DUFHORQD8QLYHUVLWDW$XWzQRPDGH%DUFHORQD6HUYHLGH3XEOLFDFLRQV
ALONSO RAMOS,0/DV)XQFLRQHV/p[LFDVHQHOPRGHOROH[LFRJUi¿FRGH,
0HO¶þuk. 7HVHGHGRXWRUDPHQWR0DGULG81('
ARAÚJO,6$OPHLGD--6LP}HV$'LDV,$SUHVHQWDomRGRSURMHFWR3HU
)LGH3DUDOHOL]DQGRRSRUWXJXrVFRPVHLVRXWUDVOtQJXDV/LQJXDPiWLFD
CALDERÓN CAMPOS,0 Sobre la elaboración de diccionarios monolingües de
SURGXFFLyQ/DVGH¿QLFLRQHVORVHMHPSORV\ODVFRORFDFLRQHVOp[LFDV*UDQDGD
8QLYHUVLGDGGH*UDQDGD
CORBÍ-BELLOT,$0)RUFDGD0/2UWL]5RMDV63pUH]2UWL]-$6iQFKH]5DPtUH]
*6iQFKH]0DUWtQH])et al. $QRSHQVRXUFHVKDOORZWUDQVIHUPDFKLQH
WUDQVODWLRQHQJLQHIRUWKHURPDQFHODQJXDJHVRI6SDLQ3URFHHGLQJVRIWKH(XURSHDQ
$VVRFLDWLRQIRU0DFKLQH7UDQVODWLRQWK$QQXDO&RQIHUHQFHSSBudapest:
(XURSHDQ$VVRFLDWLRQIRU0DFKLQH7UDQVODWLRQ
CORPAS PASTOR,*8QHVWXGLRSDUDOHORGHORVVLVWHPDVIUDVHROyJLFRVGHOLQJOpV
\GHOHVSDxRO7HVHGHGRXWRUDPHQWR8QLYHUVLGDG&RPSOXWHQVHGH0DGULG
>PLFUR¿FKDV@0iODJD8QLYHUVLGDGGH0iODJD6HUYLFLRGHSXEOLFDFLRQHVHLQWHUFDPELRFLHQWt¿FR
COSERIU,(3ULQFLSLRVGHVHPiQWLFDHVWUXFWXUDO0DGULG*UHGRV
COWIE,$30DFNLQ50F&DLJ,52[IRUG'LFWLRQDU\RI&XUUHQW,GLRPDWLF
(QJOLVKYRO,,,*HQHUDO,QWURGXFWLRQ2[IRUG283
ETTINGER,6)RUPDFLyQGHSDODEUDV\IUDVHRORJtDHQODOH[LFRJUDItD,Q+DHQVFK
*:ROI/(WWLQJHU6:HUQHU5(GV/DOH[LFRJUDItDDe la lingüística
WHyULFDDODOH[LFRJUDItDSUiFWLFDSS0DGULG*UHGRV
Extraction of restricted lexical combinations by detecting non-compositionality of multiword expressions
5
GUINOVART,;*6LP}HV$Parallel corpus-based bilingual terminology
H[WUDFWLRQ,Q/¶+RPPH0&6]XOPDQ6(GVWK,QWHUQDWLRQDO&RQIHUHQFH
RQ7HUPLQRORJ\DQG$UWL¿FLDO,QWHOOLJHQFH7RXORXVH)UDQFH1RYHPEHU
JOHNSON,,0DFSKDLO$,$7(±,QWHU$JHQF\7HUPLQRORJ\([FKDQJH
'HYHORSPHQWRID6LQJOH&HQWUDO7HUPLQRORJ\'DWDEDVHIRUWKH,QVWLWXWLRQVDQG
Agencies of the European Union. ,Q&KRL.6(G3URFHHGLQJVRIWKH:RUNVKRS
RQ7HUPLQRORJ\UHVRXUFHVDQGFRPSXWDWLRQ/5(&&RQIHUHQFH$WKqQHV*UqFH
LI,/6SRUOHGHU&&ODVVL¿HU&RPELQDWLRQIRU&RQWH[WXDO,GLRP'HWHFWLRQ
:LWKRXW/DEHOOHG'DWD,Q.RHKQ30LKDOFHD5(GV3URFHHGLQJVRIWKH
&RQIHUHQFHRQ(PSLULFDO0HWKRGVLQ1DWXUDO/DQJXDJH3URFHVVLQJSS
6LQJDSRUH$VVRFLDWLRQIRU&RPSXWDWLRQDO/LQJXLVWLFV
LI,/6SRUOHGHU&8VLQJ*DXVVLDQ0L[WXUHPRGHOVWR'HWHFW)LJXUDWLYH
/DQJXDJH LQ &RQWH[W ,Q +XPDQ /DQJXDJH7HFKQRORJLHV7KH $QQXDO
&RQIHUHQFHRIWKH1RUWK$PHULFDQ&KDSWHURIWKH$VVRFLDWLRQIRU&RPSXWDWLRQDO
/LQJXLVWLFVSS/RV$QJHOHV&DOLIRUQLD$VVRFLDWLRQIRU&RPSXWDWLRQDO
/LQJXLVWLFV
MEL’ýUK,,3KUDVHPHVLQ/DQJXDJHDQG3KUDVHRORJ\LQ/LQJXLVWLFV,Q(YHUDHUW
0YDQ'HU/LQGHQ(-6FKHDN$6FK]HQGHU(GV,GLRPV6WUXFWXUDODQG
3V\FKRORJLFDO3HUVSHFWLYHVSS+LOOVGDOH1HZ-HUVH\+RYH8.
/DZUHQFH(UOEDXP$VVRFLDWHV
MEL’ýUK,,&ODV$3ROJXqUH$,QWURGXFWLRQjOD/H[LFRORJLH([SOLFDWLYH
et Combinatoire. /RXYDLQOD1HXYH'XFXORW
MUZNY,*=HWWOHPR\HU/6$XWRPDWLF,GLRP,GHQWL¿FDWLRQLQ:LNWLRQDU\,Q
WKH3URFHHGLQJVRIWKH&RQIHUHQFHRQ(PSLULFDO0HWKRGVLQ1DWXUDO/DQJXDJH
3URFHVVLQJ(01/3SS6HDWWOH:DVKLQJWRQ86$
PADRÓ,/6WDQLORYVN\()UHH/LQJ7RZDUGV:LGHU0XOWLOLQJXDOLW\. ,Q 3URFHHGLQJV
RIWKH/DQJXDJH5HVRXUFHVDQG(YDOXDWLRQ&RQIHUHQFH/5(&(/5$,VWDQEXO
7XUNH\
PAILLARD,0&RWH[WHFROORFDWLRQVOH[LTXH,Q*XLPLHU(G&RWH[WHHWFDOFXO
GXVHQV$FWHVGHODWDEOHURQGHWHQXHj&DHQOHVHWIpYULHUSS
&DHQ3UHVVHV8QLYHUVLWDLUHVGH&DHQ
PAVEL,3$Q([WHQVLYH(PSLULFDO6WXG\RI&ROORFDWLRQ([WUDFWLRQ0HWKRGV,Q
3URFHHGLQJVRIWKH$&/6WXGHQW5HVHDUFK:RUNVKRSSS6WURXGVEXUJ3$
86$$VVRFLDWLRQIRU&RPSXWDWLRQDO/LQJXLVWLFV
SERETAN,9:HKUOL($FFXUDWH&ROORFDWLRQ([WUDFWLRQ8VLQJD0XOWLOLQJXDO
3DUVHU,Q3URFHHGLQJVRIWKHVW,QWHUQDWLRQDO&RQIHUHQFHRQ&RPSXWDWLRQDO
/LQJXLVWLFVDQGWKHWK$QQXDO0HHWLQJRIWKH$VVRFLDWLRQIRU&RPSXWDWLRQDO
/LQJXLVWLFVSS6WURXGVEXUJ3$86$$VVRFLDWLRQIRU&RPSXWDWLRQDO
/LQJXLVWLFV
5'
Joana Veloso / Alberto Simões / Álvaro Iriarte S.
SIMÕES,$$OPHLGD--1$7RROVDVWDWLVWLFDOZRUGDOLJQHUZRUNEHQFK3URFHVDPLHQWR
GHO/HQJXDMH1DWXUDO
SIMÕES,$$OPHLGD--5DPRV&DUYDOKR1'H¿QLQJDSUREDELOLVWLFWUDQVODWLRQ
GLFWLRQDULHVDOJHEUD,Q&RUUHLD/5HLV/3&DVFDOKR-*RPHV/*XHUUD+
&DUGRVR3(GV;9,3RUWXJXHVH&RQIHUHQFHRQ$UWL¿FLDO,QWHOOLJHQFH±(3,$
SS$QJUDGR+HURLVPR3RUWXJDO
SIMÕES,$&DUYDOKR1'HVHQYROYLPHQWRGHDSOLFDo}HVHP3HUOFRP)UHH/LQJ
/LQJXDPiWLFD
TOMASZCZYK,-2Q%LOLQJXDO'LFWLRQDULHV7KH&DVHIRU%LOLQJXDO'LFWLRQDULHV
IRU)RUHLJQ/DQJXDJH/HDUQHUV,Q+DUWPDQQ(G/H[LFRJUDSK\3ULQFLSOHVDQG
3UDFWLFHSS/RQGRQ$FDGHPLF3UHVV
>5HFHELGRHPGHIHYHUHLURGHHDFHLWHSDUDSXEOLFDomRHPGHDEULO
GH@