Transcript Presentation - Lancaster University

Translation corpora
and the quest for
Translation Universals
UCCTS
29.07. 2010
Anna Mauranen
Search for Translation Universals
 Characteristics that translations generally have
 began in the early / mid -1990s
 roots in translation studies and corpus linguistics
 Toury,
Klaudy
Blum-Kulka
 Baker,
Laviosa
Olohan
Why ”Universals”?
Objections from Translation Studies
”Translations inextricably linked to their particular contexts”
Any science seeks general laws, why not Translation Studies
(Chesterman)
”Impossible to capture translations from all times and all
languages” (e.g. Tymoczko)
What discipline has such access?
Not all translations are typical; Borderline cases of blends,
shortened versions etc. (Paloposki)
Translations can import new genres to cultures,
thus precede spontaneous texts in the target language/culture
- not all specimens are typical, let alone ’pure’,
why not take on the reality rather than deplore the absence of
purity?
“Talk rather about ‘laws’ or ‘tendencies’ (Toury)
Just a watered-down version of the same?
Universals are absolute, translation is probabilistic” (Frawley)
Are universals absolute?
The difference? Cf. language universals:
“Language universals are by their very nature summary
statements about characteristics or tendencies shared by all
human speakers.” (Greenberg et al. 1966)
“...universal features of translation, that is features which typically
occur in translated texts rather than original utterances and
which are not the result of interference from specific linguistic
systems.” (Baker 1993)
Universals not just linguistic features
A variety of ‘universals’ suggestions in linguistics
E.g. Bybee (2003):
“...the true language universals are universals of change.”
Most TU hypotheses phrased in process terms, as shifts;
In translation, the processes involved may be the most interesting
candidates,
or, the nature of translation as
a particular kind of language contact.
Not an exclusive focus
 The quest for universals is not the only ’core’ issue in
understanding translation.
 Others:
Typology
Variation
Change
So, why?
 Theoretical interest: what is translation?
 Descriptive interest: what are translations like?
 Applied interest: can we improve translations and translator
education with a deeper understanding of what translations tend
to have in common?
Data for universals research
 From differently related languages:
- typologically and genealogically distant
- with closer typological fit
 Different kinds of corpora
Corpus types
Bi- /multilingual corpora
Parallel corpus
Texts
and their translations
(one or multiple)
Comparable corpus
Matched texts in the
same language:
translated and
‘original’/‘spontaneous’
Matched L1 and L2 texts
(no translation)
Hypotheses on Translation Universals
Early hypotheses based on small-scale studies,
more recent on large-scale corpus studies
Most studied
‘explicitation’,
‘simplification’,‘conventionalization/normalization’;
‘source language interference’
More recent
‘underrepresentation of unique target language items’,
‘untypical collocations’
Explicitation
The most widely accepted hypothesis,
much support, little counterevidence
Translations more explicit than source texts,
i.e. the translation process tends to add information and linguistic
elements – verbalise more
Observed at different levels
(syntax, lexis, text)
Finnish > English (Parallel corpus, FECCS)
Puolueen johto oli sopinut Kekkosen miehenä tunnetun entisen ulko- ja
pääministerin tohtori Ahti Karjalaisen ehdokkuudesta ja puolueen
eduskuntaryhmän enemmistö tuki häntä.
’had agreed on ... Karjalainen’s candidacy’
The party leadership had already agreed among themselves that a known
Kekkonen follower, former foreign minister and prime minister Ahti
Karjalainen, should be their candidate.
 Syntactic explicitness, e.g. degree of ‘sentence-likeness’ increases (nonfinite>finite constructions) (cf. also Eskola 2004)
Explicitation
found also in other kinds of language contact,
e.g. lingua franca use
Simplification
Controversial; findings conflicting
Simplification at one level may increase complexity at another.
E.g. simple main clauses
may cause complexity at text level,
reducing coherent textual flow,
making it fragmented and hard to follow.
Studies on comparable corpora
The first corpus study supported lexical simplification (LaviosaBraithwaite 1996):
Most frequent lexis even more frequent in translations,
[But no less lexical variation (type/token ratio)]
Studies on CTF
(comparable Corpus of Translational Finnish, 10 million wds)
Support
Nevalainen (2005) (CTF)
Tirkkonen-Condit (2005) (CTF)
translations have more repeated n-grams:
ihan niin kuin, aivan niin kuin; samalta kuin ennenkin…
No support
Jantunen (2004, 2005) - lexis (CTF)
Eskola (2004) – syntax (CTF)
Example: degree modifiers
Jantunen (2004) : synonymous degree modifiers (hyvin, oikein, kovin)
E.g. major collocates of hyvin (Comparable corpus, CTF)
1. Original Finnish
adjectives: väsynyt, pieni
adverbs: hiljaa, hyvin, hitaasti, pian, varovasti
2. Translated Finnish
adjectives: erikoinen, hieno, kaunis, lyhyt, nuori, pieni, sairas, suuri,
tyytyväinen, tärkeä, vaalea, vaarallinen, vaatimaton, vahva, vaikea,
vakava, väsynyt, yksinkertainen, ylpeä
adverbs: harvoin, hitaasti, hyvin, kauas, korkealla, lähellä, nopeasti,
pian, pitkään, selvästi, vakavasti, varhain, varovasti
 more variation in translations
Simultaneous
simplification of lexis as overall frequencies
proliferation of variety
Example: verb frequencies
Mauranen 2000 (CTF)
e.g. Finnish verb HALUTA
HALUTA, academic texts
Original Finnish
Transl from English
Transl from other lgs
46 /mio w
101 / mio w
110 / mio w
HALUTA, popular non-fiction
Original Finnish
Transl from English
19 / mio w
31 / mio w
Example: verb collocations
HALUTA
Original Finnish:
commonest collocate KOROSTAA (‘emphasise’)
nearly 40% of all collocations
...moniaineksisuus ei ole ainoa asia jota haluan korostaa,
‘heterogeneity is not the only thing I want to emphasise’
Translated Finnish:
KOROSTAA
less than 8% of all collocations
even though HALUTA itself was more than twice as frequent
Instead, strongest collocate of HALUTA in translations:
OSOITTAA (‘show, prove’),
Tämän ainakin halusin tässä varsin luonnosmaisessa
todistelussani osoittaa
‘this at least I wanted to show in this very sketchy proof’.
But OSOITTAA never co-occurred with HALUTA in Finnish
originals
where OSOITTAA collocates with PYRKIÄ (’try’)
Koko järjestelmä on turha, kuten olen pyrkinyt osoittamaan.
’the whole system is unnecessary, as I have tried to show’
Are these findings incompatible with the “overrepresentation” of
the most frequent words?
Not necessarily:
items participating in the collocations
may be very frequent
if considered individually
 Simplification more complex than first meets the eye
 Postulate untypical collocations as a hypothetical universal
(also supported by Jantunen 2004 and Kemppanen 2008)
Untypical collocations and
unusually high proportion of very common words
also found in learner language and lingua franca speech
Simultaneous simplification of lexis (as overall frequencies)
and proliferation of variety
also in lingua franca speech
Transfer /Interference
Baker’s definition excluded interference
Earlier, Toury had formulated a
“law of interference” :
“in translation, phenomena pertaining to the make-up of the
source text tend to be transferred to the target text.” (Toury
1995)
More recently, transfer has resurfaced as a potential translation
universal
E.g.
Eskola (2004) on the basis of syntactic research (comparable
corpus, CTF)
Mauranen (2004) on the basis of lexis (comparable corpus,
CTF)
Also Teich (2003) “shining-through” (?)
English and Russian Translations compared to
Mixed Source Languages and Original Finnish
(Mauranen 2004)
Frequency bands based on rank order (Comparable Corpus of Translational
Finnish, 10 million wds)
Difference from the reference database:
Vs. Mixed-source Translations
Freq.
Eng
Russ
S
Band
1-30
63
71
134
50-79 190
115
305
100-129 104
51
155
S
357
237
594
vs. Finnish Originals
Eng
Russ
S
75
87
167
329
96
178
77
351
171
265
244
680
 Translations from different source languages had different
profiles
but
 Translations differed from originals more than from other
translations
 Transfer looks plausible
but the remaining variation must have
other explanations
What Transfer?
SLA research:
Translation studies:
transfer from L1 affects L2
transfer from L2 affects L1
Recent SLA research: L2 influences L1 (Cook 2003);
L2 learners have better L1 skills than monolinguals (Kecskes &
Papp 2000)
Transfer ubiquitous (Jarvis & Pavlenko 2007)
Translation studies: SL / ST influences TL /TT?
Optional vs. obligatory: personal pronouns to and
from Finnish
In Finnish
person reference
either by verb inflection alone
or by a combination of pronoun and inflected verb
Verb inflection obligatory, pronoun optional.
Translators often use inflected verb alone
(i.e. ‘drop pronouns’)
“ I was going to wait until another time we met, but I may as well
tell you now. I've decided to marry you.” (EO)
– Ajattelin säästää sen johonkin myöhempään kertaan, mutta
voin yhtä hyvin kertoa sen nytkin. Olen päättänyt mennä
naimisiin sinun kanssasi. (FT)
But even more often they opt for pronouns.
Translations of I, ich and minä
Two-way parallel corpus Finnish – English
English – Finnish
I
→
10742 →
I
←
5518 ←
minä
3763
minä
1471
(2.9 : 1)
(4.1 : 1)
Two-way parallel corpus Finnish – German
German - Finnish
ich →
2315 →
Ich ←
3850 ←
minä
1393
minä
942
(1.7 : 1)
(4.1 : 1)
(Mauranen & Tiittula 2005)
In sum,
translations tend to translate pronouns in the source text
 This would support text interference
Translations also reduce or add pronouns depending on the target
language
 This would support working at the level of language
Unique items
Tirkkonen-Condit (2000, 2004):
linguistic features unique to the target language
(“untranslatables”)
proportionally underrepresented in translations.
Verbs of sufficiency
Tirkkonen-Condit: Finnish verbs with the semantic feature
’sufficiency’ (Comparable corpus, CTF)
EHTIÄ (‘have enough time’, ‘be early enough’),
JAKSAA (‘be strong enough’),
MALTTAA (‘be patient enough’),
USKALTAA (‘have enough courage’),
VIITSIÄ (‘have enough initiative or energy’)
and pragmatic clitics (-kin/-kaan, -han/hän)
All proportionally more frequent in Finnish originals than in
translations.
Generic person
Similarly the Finnish ‘zero person’,
i.e. 3.person verb with no pronoun and generic meaning:
Ei tarvitse sanoa.
You don't have to say it.
‘there’s no need to say it’
(FO)
(ET)
For generic meaning,
translators tend to use more pronouns where original Finnish
employs the zero person (Mauranen & Tiittula 2005)
Unique lexical items: keli, kinos and hanki
Kujamäki (2004): text first translated into German and English,
Then students translated into Finnish (experimental study)
…lumi muuttui rännäksi ja keli vain paheni… tien viereen jäi jo
matalia kinoksia. …pian löysin itseni ja autoni hangesta.
…conditions… / ..die Strassenverhältnisse…
…a low snowbank…/…ansehnlichen Häufchen…
…in a snowdrift… / im Schnee…
keli - die Strassenverhältnisse/ conditions
36
tie/ liikenne/ajo-olosuhteet, katujen/teiden kunto, tiet, sääolot…
25
keliolosuhteet, ajokeli, keli
11
kinos - den Schnee… Häufchen/ snowbank
36
(lunta)…kasoiksi/-hin, töyräiksi, penkoiksi, tienreunaan; lumikasat…
23
lumikinoksiksi; (lunta)…kinoksiksi, lumikinos, kinosti lunta
13
hanki - ...im Schnee/ …stuck in a snowdrift
36
lumen …keskellä, saartamana, ympäröimänä;
keskellä …lumipenkkaa/-kasaa/-sohjoa/-kinosta;…
23
keskellä lumihankea; lumihangessa
13
 Underrepresentation of TL unique items – simplification or
something else?
 Would seem to suggest some sort of suppression of the TL –
even though it’s the translator’s “best” language
Conclusion
Three important things:
- Data
- Language contact
- Cross-linguistic influence
Data
Different kinds of corpora
and a broad range of languages (also non-IE)
bring out regularity and variation in translation
Language Contact
Translation universals deepen our understanding of language
contact
Shared features:
Translation, L2 learning and L2 use
- untypical collocations
- very high proportion of commonest words
Translation and lingua franca communication
- enhanced explicitness
- simultaneous simplification and increased variety in lexis
Language contact leads to cross-linguistic influence
Cross-linguistic influence
 Translation is bilingual processing;
It seems to suppress some processes and activate others
compared to monolingual processing
- activates rare collocates and rare syntactic structures
- suppresses TL-specific phenomena
(‘unique items’)
 Transfer /interference /shining-through
highly plausible even if not the whole story
 Cross-linguistic influence takes many forms and is omnipresent
(Jarvis & Pavlenko 2007)
 Translation studies: SL / ST influences TL /TT?
In all:
 Translations share many typical features,
but they are neither simple nor pure
 Much remains to be discovered about the product and the
processes