Tagungsnotizen zur Translation in Transition 4

von Mario Bisiada
Beitragsbild: Photo by Payton Tuttle on Unsplash

Vom 18. − 22. September 2019 fand die vierte Translation in Transition-Konferenz, organisiert von Mario Bisiada [@MBisiada] an der Universitat Pompeu Fabra, statt. Die Konferenzreihe wurde 2013 in Kopenhagen gegründet und fand dann in zweijährigem Abstand in Germersheim, Gent und zuletzt Barcelona statt. Die fünfte Ausgabe führt zum ersten Mal vom europäischen Kontinent weg und wird von Michael Carl und Kollegen an der Kent State University in Ohio organisiert. Eine Fortführung der Konferenz ist für 2022 in Prag geplant.

Die Translation in Transition-Reihe widmet sich der empirischen Erforschung von Translation. Der Schwerpunkt der vierten Ausgabe lag auf korpusbasierter Diskurs- und Translationsforschung. Diese Methode erfreut sich seit geraumer Zeit großer Beliebtheit und die Veröffentlichung einiger Monographien zur Lage dieser Methode zeugt von der Etablierung der korpusbasierten Forschung in der Translations- sowie Diskursforschung. In diesen Konferenzabschnitten erfuhren die Tagungsteilnehmer u. a. mehr über die Darstellung des Katalonienkonflikts in der deutschen Presse, über verschiedene Framings von #MeToo auf Twitter im Englischen, Spanischen und Deutschen oder die Lokalisierung von Fremdheit in russischen und chinesischen Untertitelungen.

Die Tagung umfasste aber auch weitere korpusbasierte Forschungsthemen wie auch Umfrageforschung, Lehrforschung oder prozessbasierte Forschung. Aufgrund des aktuellen Digitalisierungsschwungs stand das Thema der Evaluation von maschinellen Übersetzungen und Post-Editionen ebenfalls weit oben auf der Agenda. Als Dauerbrenner erweisen sich auch Studien zu verschiedenen Translationsmodi und damit verbundenem kognitiven Aufwand.

Tagungsnotizen

(Kollaboration von Oliver Czulo, Clara Ginovart, Felix Hoberg, Katja Lapshinova-Koltunski & Éric Poirier)

Donnerstag, 19. September 2019

Plenary 1

Maeve Olohan: Studying translation through its doings and sayings

  • „Practice theory“: social fields constituted by practice, using specific competences, working with materials, intending meanings; i. e. an integrated view on dimensions that are often studied separately in Translation Studies
  • practice is already ‚there‘ (historically shaped), the person is the carrier of the practice (and may change it) -> focus on the practice, not on the person; challenge: do not re-ify ‚practice‘!
  • ethnographic (or rather praxicographic?) approach
    • e. g. with the interview to the double: what would we tell a double so it wouldn’t be recognised as such (rather than a high-end account of how sth should be done)
    • or a ‚guided tour‘ drafted by the interviewee
  • quantitative data to complement qualitative approach? matter of debate
  • interdependent practices: how are they set apart?
  • dialectical approach for co-evolving practices such as MT and human translation: tensions and contradictions between such practices
  • mixing practices: in the translation classroom, practice of translating und practice of learning are mixed and sometimes it is hard, though necessary, to understand which one of the two is at work in the given moment

Parallel session 1

Ekaterina Lapshinova-Koltunski: Analysing the dimension of Mode in translation

  • languages vary according to their context, we can trace this variation using features from variational linguistcs (SFL and genre/register studies), also variation between spoken and written (Mode)
  • can we also trace this variation (Mode) in Translation (translation+interpreting)?
  • which features vary between the two Modes interpreting and translating, and in non-translated language (Do these features coincide?)?
  • focus corpus: interpretations from Europarl (EPIC-UdS), CroCo (prof.translations into German and German non-translations), VARTRA (student translations of the English texts from CroCo) and GECCo (spoken registers in German)
  • the common features which are informative for Mode distinction in both translated and non-translated German: general nouns and modal verbs of obligation. Buth others have a certain overlap: pronouns, conjunctions
  • the features are good in classifying mode (in both translations and non-translations), but not so good in differentiating student vs. professional translations
  • classifiers work considerably well, but problem is the small amount of data on the interpreting side, probibiting creation of better classification models

Eric Poirier: Content-based and quantitative analysis of speech and discourse translations to detect translation shifts in target texts and meaning discrepancies between two target texts

  • New title suggested: Quantitative analysis of political discourse translations to detect translation shifts related to information additions and omissions
  • Description of a Python algorithm (version 2.0) created and used for detecting translation shifts in parallel corpora of political speeches in English, French and Spanish
  • Natural language processing of parallel segments based on open-source Spacy modules (spacy.io)
  • Translation shifts were detected based on information addition or omission and classified between free (non-mandatory) shifts and fixed (mandatory) shifts
  • Information additions and omissions were measured for three language features: character length, word count and lexical word count
  • Different target-to-source segment ratios were tested with the three features, with more specific attention to lexical word count ratios or information precision ratio
  • Corpora analyzed as for now:
    • 6.5 million words for 515 Québec laws translated from French to English (paper in preparation with Jean-Hugues Roy)
    • 8,432 words for Theresa May and J F Kennedy speeches translated from English to French (paper on version 1.0 algorithm to be published in next issue of TTR)
    • 13,802 words for A. Lincoln, J.F. Kennedy, M.L. King, B. Obama and D. Trump speeches translated from English to Spanish
  • Results:
    • Very strong correlations between source and target segments for each 3 features with similar correlation strengths for each language pair
    • Character length correlation was stronger than lexical word count correlation which was stronger than word count correlation in all three translation directions (French to English, English to French and English to Spanish)
    • Best metric for addition and omission detection in all three languages was standardized length Euclidian distance in 3-dimensions (character length, word count and lexical word count)
  • Some testing need to be done for 2-dimensions Euclidian distance such as character length and lexical word count (which have the strongest correlation with segment length)
  • Example of a shift detected in the Obama speech to Cuba: „Not because American Democracy is perfect, but precisely because we’re not. = No porque pienso que la democracia en Estados Unidos sea perfecta, sino precismente porque no lo somos.
  • Numerous translation shifts were found in the 6.5 million words law corpus but much less in the other political discourse corpora
  • Two possible reasons: political discourse corpora being very small compared to law corpora; also perhaps the strategic nature of texts of the political discourses which favour isomorphic translation techniques over heteromorphic (shifts) translations

Isabela Lacruz, Haruka Ogawa & Michael Carl: Indicators of cognitive effort in translation from Japanese into English and Spanish

  • HTra as measure: Relative frequency of the variation in the translation of a certain segment; here used as measure for cognitive effort
  • HTra rends to be higher when target language is „far away“ from source language⁰
  • two texts translated into Japanese and Spanish; translators asked to assess whether they thought the translation would be hard or easy for them
  • groups were set together according to L1 and L2, e.g. L1 Japanese, L2 Spanish
  • high HTra for certain (often formulaic) expressions in target language such as „es decir“
  • other factor differences in culture where explicitation or explanation into one language is more prevalent than into another; implicitation seems to be more often the case into Spanish
  • combinations such as L1J into Spanish yield higher HTra than into English, though not necessarily on a statistically different level

Parallel Session 2

Marcello Giugliano: Images of the Catalan national identity in the German Press: A corpus-based approach

  • how does German press represent Catalan conflict? Which role does translation play in there?
  • Media translation relatively new; problem: often, we don’t have a concrete text, but rather a narrative that is translated
  • exploratory corpus-based design
  • early September to early October 2017: reports from mainstream newspapers in Germany on Catalonia conflict (CAT), and two Catalan outlets (CATRef)
  • keyword analysis of German np: two lists of keywords which suggest opposition, e.g. Madrid and Barcelona, but also conceptualisations such as independence vs. secession
  • Fillmorian frame semantic approach: hyperframe „geopolitical separation“ where different terms put a different focus on agents with diverging assessments of their role
  • funnily, two ideological opposed nps in Germany use similar discursive strategies, leaning towards Rajoy pole – maybe a divergence in frame connotation between German and Spanish? or due to information sources which favour one position? -> further research

Mario Bisiada: Already a movement or still a debate? Framings of the #MeToo hashtag in German, English and Spanish

  • functions of hashtags: categorisation, but also evaluative potential, sometimes linguistic innovation, paralanguage that may be more powerful than emojis (meta communication)
  • translating hashtag defeats the purpose of cross-border solidarity (e.g. #metoo to #moiaussi): potentially decreases the audience
  • various studies found that Twitter is often used to channel emotions, but not so much to engage in dialogue, so a lot of people talking about the same topic
  • #MeToo already coined in 2006, but only took on in 2017: encouraged members to showcase magnitude of the problem of sexual violence
  • what is the semantic prosody and the evaluative stance of the authors?
  • tweets from late July and early August 2019, around 450 tweets per language
  • English frames #MeToo as ‚movement‘ and discusses specific cases
  • Spanish also talks about ‚movimiento‘
  • German talks about a debate and hysteria, so rather meta-talk than concrete to the topic
  • cross-linguistic framing: #MeToo as organised pressure group (hysteria, false accusation etc.); as exaggerated (rebellion, fetish); questioning the veracity (nonsense, nutcase, house of cards)
  • an interesting category is hijacking esp. in Germany, in order to promote other ideologies, e. g. „all these foreigners rape women #metoo“
  • in German, #MeToo is framed more negatively than in other languages

Plenary 2

Josep Marco: Weaving translation properties together through the gravitational pull hypothesis

  • problem: Translation specific property of bith over- und underrepresentation of certain features seems contradictory; prototypicality will lead to overrep, connectivity of related concepts to both over- and underrepresentation
  • current version of the theory differentiates various factory:, gravitational pull pertains to SL, magnetism to TL, there is also connectivity which links the two
  • Cognitive Grammar (Langacker) and theories of the bilingual mind (esp. De Groot) as pillars for the hypothesis
  • Cognitive salience and Entrenchment as cognitive basis
  • appeal: move-away of binary relations (multifactorial), rooted in cognition, predictive/explanatory power
  • study on diminutives EN-Cat, DE-Cat
    • Unique item in EN-Cat, as suffix-diminutives are rare in English, unlike in German
    • hypothesis: over-represented in DE-Cat
    • procedures: retrieved with the help of CQP and filtered
    • result: more frequent in DE-Cat, but not more than in CatOrig
    • maybe diminutives are not so salient in German or Catalan; patterns of connectivity should be revisited?
  • study on ‚caldre‘ in Catalan which has no direct equivalent in English (unique item in EN-Cat), but good structural equivalent in French with ‚fallloir‘
    • hypothesis: under-representation in EN-Cat, overrepresentation in FR-Cat
    • result: confirmed
    • possible explanation: ‚caldre‘ not so salient in Catalan, so no magnetism
    • there are some strong triggers for caldre when seen from TL side, but when seen from SL, these triggers mostly result in different translations such as ’necesitar‘
    • calculating mutual correspondence (MC) between translation equivalents (e.g. caldre and need)
    • connectivity of e.g. must/haver de stronger than for must/caldre, lending itself as explanation for under-rather ghan over-representation
    • quite the opposite for FR-Cat, where falloir and caldre are very strongly connected
    • haver de and have to are cognates, falloir and caldre are very simliar in syntactic structure -> may facilitate connectivity
  • Conclusions:
    • GPH has high potential, but is still underexplored; need better understanding of how the three factors interact
    • testing out GPH is time-consuming as it needs other data than only parallel(?) corpus!

Parallel session 3

Maria Kunilovskaya & Ekaterina Lapshinova-Koltunski: Lexicogrammatic specifity of out-of-English translations to German and Russian

  • Translation pairs EN-DE, EN-RU; varieties: professional, student
  • Background: techniques in Translationese studies and Register variation studies; Translation competence: level of professional expertise can be linked to translationese
  • Question: How do lang. pair and variety compare wrt strength and types of translationese?
  • EN-DE: CroCo, VARTRA; EN-RU: RusLTC, RNC
  • other sources are hard to find, as you want register variation and student translations
  • translationese features, mostly from Evert/Neumann
  • Pre-Test: do these features capture translationese?
    • SVM classification into trans and non-trans
    • EN-DE: F1 = 0.79, EN-RU: F1 = 0.91
    • translation status and varieties distribute differently in a Principle Component Analyses (PCA), where translations EN-DE shift towards English through shining through
  • Experiment:
    • Euclidean distance: EN-RU (1.05) is not so much further away then in EN-DE (1.05)
    • for EN-DE, shining through is stronger (0.922) than in EN-RU (0.7xx)
    • when adding the varieties pro vs. stu as variables, results are mixed; e.g. EN-RU, professionals produce more independent translationese effects, but students stronger shining through(?)
  • Conclusions:
    • different feature sets for shining through and independent translationese effects
    • more shining through in EN-RU than EN-DE
    • varieties differ in amount and type of translationese, with students yielding an unclear picture
    • potential explanation for differences for the latter: professionals have a routine and thus reproduce certain features more prominently

Alina Karakanta & Elke Teich: Detecting and analysing translationese with probabilistic language models

  • Translationese as ‚finger print‘ of a mediated communication event
  • in Corpus-based translation studies: a set of features tested for significance; in Computational linguistics: as text classification task
  • probabilistic models shall uncover relevant features, help classify kinds of effects (e.g. simplification, normalization) and find the loci of effects
  • Europarl-UdS corpus: EN-DE (English-German), where the source side is English original (usually, source languages are mixed in the minutes of Europarl)
  • relevant features
    • Relative Entropy
    • DE-Original: impersonal structues, pronouns, deictics; EN-DE-Translations: personal structures, pronouns
  • kinds of effects
    • Perplexity
    • based on POS-tagged translation corpora
    • variability of translation of POS patterns(?)
  • loci of the effect
    • Probabilistic Language Model
    • even inside a sentence, it can be detected what is unusual in terms of probability of occurence
  • how does this compare to human assessment?
    • 2 advanced students asked to state whether they believe a sentence is an original or translations, some with strong, some with little Perplexity e
    • accuracy is very low, but it seems that students very much focus on the lexis, interfering with the task; expectation: translations long and complex, originals simple and understandable
    • but: when asked to mark parts which seem to be an indicator for a translation, they often marked parts which were also marked by the probabilistic language model as unusual
  • Conclusion:
    • Relative entropy good to detect useful features, robust to low frequency
    • Perplexity food to detect interferences and normalization

Parallel session 4

Olga Nádvorníková: Stylistic normalisation in translation: The case of the Czech transgressive

  • Czech transgressive:
    • converb which marks adverbial subrdination
    • similar to -ing adjuncts in English, gerunds in Romance languages
    • conveys e. g. accompanying circumstances
    • complex form
  • why is Czech different from other Slavic languages which use morphologicall simpler adverbial structures? Historical background with revival of Czech language in 19th century, looking back into (pre-)17th century for codification
  • nowadays, used very rarely and stylistically marked as ‚bookish‘ or ‚archaic‘
  • Hypothesis: in translated texts, should be less frequent than in non-translations, due to normalization
  • Research question: what are linguistic and extra-linguistic factors?
  • Method: corpora and questionnaires
  • Data:
    • Corpus: Intercorp, here the Core part with all the metadata (author, date of publication); mostly fiction
    • Questionnaire: 87 translators, 33 publishers
  • Frequency (wrt statistical significance):
    • genre: transgressive more frequent in fiction than non-fiction
    • status: transgressive more frequent in non-translated than translated for fiction, but not for non-fiction!
    • form: present transgr. much more frequent than past transgr.
  • Factors:
    • date of publication (unsurprisingly)
    • sex of the translator: men use transgressive more often! but why?
    • Publishing houses (from questionnaire): Only around 20% have an internal codified norm, some a tacit convention as to transgressive
    • Translators (from questionnaire):
      • for certainy text types such as historical text
      • but also handy in use for syntactic condensation
      • some have an emotional stance, e.g. they try to avoid transgressive in order not to sound archaic, or they simply state a love for the form
Freitag, 20. September 2019

Plenary 3

Oliver Czulo: Re-introducing Frame semantics to Translation studies

  • According to functional translation theory (Reiss & Vermeer, 1984 skopos; Holz-Manttari, 1989; House, 1977, 1997; Nord, 2006:43) translation decisions are strategic and not based (primarily) on equivalence, but how then do we assert a translation relation? Tymoczko (2005): similarity instead of identity.
  • Formal approaches to identity / similarity:
    • example: direct object in sentence initial position in German when translated to English
    • Modulations (Vinay & Darbelnet 1959) / shifts (Catford, 1965).
    • Strategies (examples): Reordering; Passivization + Verb accomodation; Topicalization; It-cleft; parallelization.
  • Functional linguistic approach:
    • Construction Grammar as a device to describe also functional aspects of such strategies (such as attention guidance, politeness)
    • conceptually, constructions are form-function-meaning pairs (meaning as propositional content)
    • The definition of function here is a weak part of this theory/framework/assumption. Constructions are non-compositional, contain lexematic units or are fully schematic.
  • Semantic approach: Frame semantics. (Fillmore 1977, 1982, 1985. Petruck, 1996). How do we understand? A frame attributes certain roles to elements and describes relations between them.
    • There is a number of studies using Frame Semantic for Translation (Snell-Hornby, 1986; Kußmaul, 2000; Pado/Erk, 2005; Rojo, 2002, Czulo, 2017).
    • A frame is a mental categorization device to which logics of action are connected. In the frame of a ‚presentation‘ you have an audience, speaker, and a medium, the speaker is speaking, with logics of action such as the audience listening, which doesn’t work outside the presentation scenario.
    • Frames are also related to another, e. g. through Precedence, some event has to occur before the other. Related frames evoke each other or can be invoked (mechanisms yet to be investigated more deeply).
    • Evocation and perspectivization: ‚the speaker was giving a talk‘ (implicit: Audience).
  • Primacy of frame model:
    • hypothesis: maximally comparable frame in the given content in the target language
    • override factors such as typological or contrastive differences between languages can lead to slight shifts in semantics
    • frames evoked should, however, often be relatable to each other closely through the frame hierarchy, allowing to state a certain semantic similarity between original and translation
  • Elevating to text level:
    • Text is a network, but also (at least partially) a linearized form of communication. There are few instances of fully annotaded texts (for frames, frame-database). EXAMPLE: ‚The guard entered the room. The alarm went off.‘ VS. ‚The alarm went off. The guard entered the room‘; here, implications between the ordering of evoked frames is important, but how do we account for this?
  • Elevating to discourse level:
    • Discourse as patterns of what can be said, what is said and who can say what
    • ‚what is said‘ can be modelled by means of Frame semantics (patterns of assigning roles and relations)
    • problem: these patterns can be transferred/copied from a language/culture without having a concrete source we could point to
  • CONCLUSIONS AND OUTLOOK: How to compare and link frames, frame relations across languages? How to model factors like linearity, argumentation structures? Is existing application to MT evaluation transferable to HT analysis/evaluation?

Parallel session 5

Madjda Chelli: Arab Muslim Anglophone Literature in the classroom

  • Cultural translation: what happens when two semiotic systems (cultures) are in contact
    • Literary translation as a use of domestic vocabulary to describe a different cultures
    • Homi Bhabha: third space, hybridity to theorize questions of identity in the process of cultural translation
    • Ribeiro: Cultural translation as means to resist colonizing tendencies of globalization
  • Arab Muslin women Anglophone Literature: e. g. women with Arab background born in the UK and writing in English
    • representing Arab women to Western readers through their lense, not through outside
    • an instance of hybrid entities
    • assuming the role of double-sided mediators
  • a minor literatur in many ways: written by migrants, women, with non-sacular and non-Christian perspective; perspective which is often rejected by Westerners
  • adapting to the classroom
    • Literature as comparable corpora: same language, different text production backgrounds
    • Antconc as tool for research on linguistic and/or cultural features in Arab Muslim Anglophone Literature
    • e.g. detect examples of representation of Self and Other, decide on cultural keywords

Leticia Tian Zhang & Ludmila Shafirova: Localizing the Other in Chinese fan subbing and Russian fan dubbing

  • Fans as cultural prosumer and members of a participatory culture
  • Ethnographic case study
    • particpant observation of dubbing in recording
    • interviews with researchers
  • Example from China; movie subbing
    • urban, well educated, altruistic young crowd
    • collaborate with mainstream video and translation companies
    • impulses also from teaching
    • Localization strategies (examples)
      • traditional Chinese proverbs: saves space, introduces literary style
      • Chinese characters with visual similarities to memes (e. g. sad face)
  • Examples from Russia; ‚Brony‘ culture
    • young men adapting pony symbolics and video material usually targeting girls
    • fandubbing ‚My little pony‘ EN-RU: using techniques from simultaneuos interpreting
    • official dubbing simplifies, for the intended target group; fan translation tries to include puns in the translation
    • fandubbers also act out the part
  • Comparison: specific audience groups instead (RU) instead of general audience (ZH); addition of meaning to specific audience (RU); fan pride as important value (RU); fun as an important emotion instead (RU) of intellectual engagement (ZH)

Madiha Kassawat: Bilingual parallel analysis of Arabic and French blurbs

  • digital environment absorbs 20-30% of advertising expenses
  • websites as culturally localised products
  • Localization as making a product linguistically and culturally appropriate to target market
  • a locale can be polyvalent in use, e. g. for Spanish it is mostly Spain, but to a similar extent in Mexico, the US, Argentina
  • internationalized text is simplified in order to make it accessible / translatable to an international market, removing ‚traces of culture‘
  • Data:
    • three Corporate websites, marketing texts (cosmetics, furniture, technology)
    • ST: Int. English, TT: Standard Arabic, French (France, Canada, Switzerland, Morocco)
  • Findings:
    • Product ‚Super Slow-mo camera‘
      • Arabic: explanatory translation of the title, with shorter description but using some literary style
    • Product ‚Shoe, coat, hack rack‘
      • French: literal translation
      • Arabic: quite literally, but the dressing code is different in many Arabic regions, you don’t wear hats, but e. g. head scarfs – the translation doesn’t account for that
    • Product ‚mobile phone‘
      • variation between French and Swiss versions; both seem to be applicable to the other country, so not country specific, but shows creative element of the localization
    • Product ‚GALifornia powder blush‘
      • product name kept both in French and Arabic
      • skin colours: pink blush in French, brown blush in Arabic
      • Arabic: ‚who doesn’t like the California girl‘ – but how should they know? building on stereotypes?
    • Conclusions:
      • internationalized text can facilitate the translation, but can introduce foreign cultural references as well

Felix Hoberg: Evaluation of MS Sykpe Translator: A dialogue-oriented conception in the Catalan-German language pair

  • challenges: specific type of communication: abbreviations, orthography, lack of non-verbal elements
  • eye-tracking study: (1) Catalan-German communication – participants (students) who needed to go to Catalunya and had a chat with Catalan native speakers (6) via Skype using MS Skype Translator and also (2) German-German communication (comparative study)
  • data: 16/20 (so far) data sets: 287 min in total, 18 minutes on average per participant –> 287 DE messages, 283 DE MT messages; 407 Catalan messages, 403 Catalan MT messages, diverging numbers due to latency of Skype Translator who inexplicably merged some messages into one
  • Results: comparative AOI results
  • Average fixation duration: fixation duration on German original messages is relatively short compared with the fixation count on all German original AOI; vice versa, AOI on MT output into German is extremly high in terms of count but only slightly higher in duration than GerO
  • Saccades: looking at saccades starting in one AOI and ending in another, most saccades happen to occur between corresponding AOI (GerO -> MTintoCat & vice versa // CatO -> MTintoGer & vice versa)

Jennifer Vardaro, Moritz Schäffer & Silvia Hansen-Schirra: Investigating post-editing behaviour at the European Commission

  • Data: En, NMT, NMTPE, REV (revised German) of PENMT, 24 texts each
  • drawback: work with sentences only
  • Error types: (1) NMT errors and (2) which errors post-editors corrected
  • Error categories by Maja Popovic (Hjerson)
  • MQM: Error category results for NMT and NMTPE (freq.) – register/style category is the most freq. one for NMT and NMTPE
  • Second study: eye-tracking – which error-categories were …?
  • Measures: early and late measures –> when were the different error categories recognised (early or late?)
  • Statistical Analysis with Linear Mixed-Effect Regression Models
  • No difference in eye movement for different error types, errors are recognised

Parallel session 6

Àlvaro Calero Pons: Editors as definers of language standards: A comparison of co-published literary translations in Catalan and French (France-Québec)

  • standardization of minorized languages: ex. of caixa and capsa
  • Harry Porter in Catalan: only first 2 books were adapted;99,7% of young Valencians normally read in Spanish
  • Catalan and French as pluricentric languages
    • Catalan: norm regionalization; pressure towards bilingualism
    • French: bilingualism in Québec, but not France
  • Catalan: Harry Potter / French: three novels by Jonathan Franzen
  • Example: Harry Potter in Catalan
    • 1st book: within the valentian version, not very coherent in which variant is enforced (e.g. sometime caixa, sometimes capsa)
    • 1st and 2nd book: variation again, e. g. between nit and vespre for evening, this time variation between Barcelona and Valentia version
    • hispanisms occur
  • Example: Franzen translations into Québec French
    • as in the case of Catalan, changes are not systematic
    • Francisms appear, such as boulot -> emploi
    • Anglicisms
  • Conclusions
    • changes due to co-publishing agreement, decided by editors
    • status as translation legitimates enforcing norms as translations are perceived to be outside the language community

Jonas Freiwald, Tatiana Serbina & Haidee Kruger: Marked themes, do editors change them? Editorial influence on Theme in English

  • focus on the translation product
    • focus on linguistic features
    • data: CroCo Corpus (Hansen-Schirra et al., 2012)
  • Also Focus on translation process (eye-tracking)
  • Lab experiment to compare translations on site to corpus data:
    • e. g. for subject agency, lab translations did not yield same results as corpus; for theme markedness, there was no difference
    • ecological validity, observer’s paradox, editing – which factor plays which role?
  • previous studies have shown that covert norms are at play, editing has influence on a number of linguistic features
  • theme
    • as point of departure of the message (SFL perspective)
    • identified positionally early in the clause: first experiential element of the clause
    • marked and unmarked themes
  • Question: Do editors change marked to unmarked themes?
  • Data:
    • parallel corpus of original writing and edited counterparts
    • South African and Australian as varieties, five text categories
    • annotation of 1400 sentence pairs with thematic structures
  • Results:
    • number of marked themes tended to increase rather than decrease in editing, both when looked from perspective of text category and varieties
    • numbers for types of themes all in all very stable and similar across text categories and varieties
    • none of the factors variety, text category or editing were a good predictor for increase in marked themes
    • in White South African English, there is a significant effect on theme changes as opposed to other varieties
    • marked themes are more likely to be changed, which is maybe due to the low number of marked themes
  • Issues:
    • the corpus was not a translation corpus
    • varieties in Kruger’s corpus are rarely found in Europe where US and British dominate

Schreibe einen Kommentar

Deine E-Mail-Adresse wird nicht veröffentlicht. Erforderliche Felder sind mit * markiert.

Diese Website verwendet Akismet, um Spam zu reduzieren. Erfahre mehr darüber, wie deine Kommentardaten verarbeitet werden.