Accessibility / Report Error

Machine Translation and Poetry. The Case of English and Portuguese

Abstract

This article sets out to analyse the translations into Portuguese of three poems written I English, “A Letter is a Joy of Earth” by Emily Dickinson, “To a Stranger” by Walt Whitman and “Sandra” by Charles Bukowski. Three translations by three human translators, Geir Campos, Pedro Gonzaga and Jorge de Sena, are compared to the translations made by Google Translate in order to evaluate machine translation quality. This research shows that machine translations are less “ludicrous” than some would think and are in fact quite acceptable. In the cases investigated, machine translations are sometimes as acceptable as the ones made by the professional translators, and they could even help them to make mistakes through a lack of attention or by ignoring all the possibilities in the case of polysemous words. The Google translations are obviously plainer and there are a number of mistakes in them of the kind one expects: wrong concordances, wrong interpretations of polysemous words, wrong interpretation of gendered words. However, overall the results are far more satisfying than forecast. Google Translate, or similar programmes, may help translators with different, albeit sound alternatives. Additionally, machine translations provide a useful tool to analyse the idiosyncrasies of translators.

Keywords:
Machine translation; Poetry; Portuguese-English

Introduction

Machine translation is an area in full expansion. Google Translate and the like might have been the laughing stock of many plurilinguals for a long time, but it is now considered a quite reliable way of translating texts, certainly when of a technical kind. Google Translate has been followed suit by Microsoft and, of lately, by DeepL. Where progress in machine translation had been slow over the last decades, new data mining techniques have brought about great progress in the last two years, especially since the emergence of Neural Networks. It goes without saying that the driving force behind this evolution is industry and trade, both relying heavily on translation to conquer the global market. The ever-increasing globalisation of trade indicates that this evolution will lead to increasingly more refined machine translation protocols, reducing the need of intervention of the human translators, their jobs being redefined as pre- and posteditors of texts translated by computer programmes.

When looking at the literature on the subject of machine translation, it is difficult to get rid of the impression that linguists and other traditional language experts are hardly being involved in this quest. A search on Google Scholar with the term 'Machine Translation' shows that the academic authors of articles on this subject are all employed at technological faculties. It is also striking that a great deal of them are Chinese, maybe an indication that the interests of researchers involved in machine translation are being stimulated primarily by global trade. Machine Translation is foremost “big business”. It is therefore not surprising that a combination of search terms such as “Machine Translation Prose”, concentrating the search on literary topics, does not yield many results. One does not expect computer engineers to be enthused by the translation of Flaubert or Cervantes. It is therefore all the more surprising that the search term “Machine Translation Poetry” does yield a whole series of results, this time again with Chinese authors as a majority. Especially Chinese computer linguists have been interested in the topic of “producing poetry” (Qixin Wang,Tianyi Luo, DongWang, Chao Xing, 2016Qixin Wang, Tianyi Luo, Dong Wang, Chao Xing. (2016). Chinese Song Iambics Generation with Neural Attention-based Model. Consulted https://arxiv.org/abs/1604.06274v2
https://arxiv.org/abs/1604.06274v2...
) (Jiyuan Zhang, Yang Feng, Dong Wang, Yang Wang, Andrew Abel, Shiyue Zhang, Andi Zhang, 2017Jiyuan Zhang, Yang Feng, Dong Wang, Yang Wang, Andrew Abel, Shiyue Zhang, Andi Zhang. (2017). Flexible and Creative Chinese Poetry Generation Using Neural Memory. Consulted https://arxiv.org/abs/1705.03773
https://arxiv.org/abs/1705.03773...
). The results are truly fascinating, even if outside the immediate scope of this article. Generating poetry through machines is not the same as translating it; it only seems more difficult and the point of coincidence is that a machine is used to deal with language in a highly creative and supposedly unexpected way.1 1 It is interesting for literary scholars to take a look at the drawbacks cited by scientists when it comes to the translation of poetry by machines. For example, [xref ref-type="bibr" rid="r3"]Ghazvininejad, Shi, Priyadarshi & Knight (2017[/xref], p. 43) cite three disadvantages. Number 3 is the next one: "Slow generation speed. Generating a poem may require a heavy search procedure. For example, the system of Ghazvininejad et al. (2016) needs 20 seconds for a four-line poem. Such slow speed is a serious bottleneck for a smooth user experience, and prevents the large-scale collection of feedback for system tuning. Certainly not a consideration that a literature scholar would make.”

As far as the translation of literary prose is concerned, there have been a few initiatives such as those reported on by Toral and Way (2018Toral, A., & Way, A. (2018). What Level of Quality Can Neural Machine Translation Attain on Literary Text? In J. Moorkens, S. Castilho, F. Gaspari, & S. Doherty (Red.), Translation Quality Assessment (Vol. 1, pp. 263-287). Cham: Springer International Publishing. https://doi.org/10.1007/978-3-319-91241-7_12
https://doi.org/10.1007/978-3-319-91241-...
) and Toral, Wieling, and Way, (2018Toral, A., Wieling, M., & Way, A. (2018). Post-editing Effort of a Novel With Statistical and Neural Machine Translation. Frontiers in Digital Humanities, 5. https://doi.org/10.3389/fdigh.2018.00009
https://doi.org/10.3389/fdigh.2018.00009...
). In “What Level of Quality can Neural Machine Translation attain on Literary Text?” Toral and Wieling set out to compare three kinds of translations of English novels into Catalan by means of traditional Machine Translation (statistical phrase-based Machine Translation PBMST), Neural Machine Translation (NMT) and humans. The study includes literary works like Harry Potter by J.K. Rowling, but also Ulysses by James Joyce. Not surprisingly the machines performed better in the case of Harry Potter than of Ulysses. The authors conclude that human translation is still the better option, but that NMT meant a great improvement when compared to PBMST and that the gap with human translation is therefore narrowing. Human translators were asked to evaluate the translations of three books done by machines and by humans and the authors conclude: "NMT outperformed PBSMT. For two out of the three books native speakers perceived NMT translations to be of equivalent quality to those of human translations in around one third of the cases (one sixth for PBSMT)” (p. 22). When it comes to translating novels, Machine translation still has a long way to go so as to attain human capacity, but these studies reveal that machine translation can at least be of great help, also to professional translators. Besides, one detail catches the eye: machine translation seems to be better at translating sentences when they are shorter, especially in the case of PBSMT. This seems like an indication that poetry might stand a chance when translated by machines.

Another article by Toral and Way (Toral et al, 2018Toral, A., Wieling, M., & Way, A. (2018). Post-editing Effort of a Novel With Statistical and Neural Machine Translation. Frontiers in Digital Humanities, 5. https://doi.org/10.3389/fdigh.2018.00009
https://doi.org/10.3389/fdigh.2018.00009...
) gives a good overview of what has been achieved in the field of machine translation and literary texts. In Toral and Way (2018) the authors analyse a chapter of a science fiction novel translated by, first, six professional translators and, second, by means of post-edited PBMT and post-edited NMT. They conclude that NMT, even more so than PBMT, helps when translating literature, at least in the case of novels, and that for professional translators the procedure is certainly faster than working purely from scratch. However, the hypothesis the authors raise is that working with machine translation can lead to the final result bearing the traces of its passage through the computer. In short, literary texts are not immediately suitable for machine translation, but methods are improving and even in the present state machine translation helps at speeding up the process.

In “Traduction automatisée d’une œuvre littéraire: une étude pilote”, Laurent Besacier states that current machine translation (MT ) techniques are continuously improving and he inquires if the “pipeline (MT+PE)” (Machine Translation with Post Editing) could be useful translating literary works? (Besacier, 2014Besacier, L. (2014). Traduction automatisée d’une oeuvre littéraire: une étude pilote. Traitement Automatique du Langage Naturel (TALN). Consulted https://hal.inria.fr/hal-01003944/document
https://hal.inria.fr/hal-01003944/docume...
) To investigate this, Besacier machine-translated a short story by the American writer Richard Powers and then had it post-edited and revised by (nonprofessional) translators. The translation output was subsequently evaluated by a panel of readers (who read the translated short story in French and answered a survey afterwards). Eventually, the translated text was submitted to Powers’ official French translator, who formulated a few very interesting comments.2 2 "Il reste bien sûr des imperfections, des lourdeurs, voire des erreurs ponctuelles, qui appellent une correction" Principales erreurs ? - "Le défaut le plus répétitif, celui dont souffre d’ailleurs le travail de tout traducteur débutant, est le calque syntaxique, là où le français structure différemment la phrase .../... On comprend, mais ça ne sonne pas vraiment français" - "Autre défaut assez fréquent, la perte des idiomatismes du français au profit d’anglicismes. (…) " - "Un troisième défaut tient à la non prise en compte de certains repères culturels .../... Par exemple, Powers fait plusieurs références à la topographie de Boston qui donnent lieu à des inexactitudes dans la traduction : « la rivière Charles » par exemple (p. 12) qui n’est pas une rivière mais plutôt un fleuve ; c’est pourquoi on traduira par « la Charles River » ou simplement « la Charles » (…)" The translated text is generally readable, and some scientific passages are very convincing, but there are clumsy wordings: "The most repetitive defect, from which the work of any beginner translator suffers, is situated at the syntactic level, since the French language structures a sentence differently. The translation is understandable but does not really sound like French." Other shortcomings are the anglicismes which he calls “Frenglish”. Finally, cultural references are (evidently) not taken into account. This summarizes well what one would expect to be the drawbacks of MT. The lack of awareness of context is probably the cause of most of the mistakes encountered.

The conclusion is that machine translation is booming business and that linguists seem to play no major role in it. The recent use of Neural Networks is revolutionising Machine Translation and progress is rapid due to the ever-growing input of translated texts. Finally, and contrary to popular belief, computer engineers and scientists are sometimes also interested in literature.

The research

All forms of machine translation are based on the premise that language repeats itself and that especially “word sequences” repeat themselves. The order in which words are used is not coincidental, neither is it the object of a series of continuous choices. It might be useful to recall the findings of John Sinclair here (1995)Sinclair, J. (1995). Corpus, concordance, collocation (3. impr). Oxford: Oxford Univ. Press.. Sinclair’s main discovery was that people do not write or read by filling separate slots one by one, but used “chunks”, sequences typically between two and five words. Instead of the “slot and filler” principle, the mechanism that would govern linguistic communication would be the “idiom” principle. Several words are selected in a row and used over and over again. The more a text is governed by this “Idiom principle”, the more predictable it is. Good examples are technical texts, journalistic texts and scientific articles. The novelty of these texts lies not in the form, but only in the message. Quite another story are literary texts, where form is just as important as the message and authors tend to avoid chunks, preferring the more laborious technique of choosing words one by one. This is why “Literature is news that stays news”, as Ezra Pound used to say.

In what follows I will analyse the translation of three poems in English by canonical authors: “A Letter is a Joy of Earth” by Emily Dickinson, “To a Stranger” by Walt Whitman and “Sandra” by Charles Bukowski. The poets were chosen because of their notoriety. From each we compared the “official” translation, published from the hand of a professional translator, with a translation done by Google Translate. The authors of the poems, Dickinson, Whitman and Bukowski, can do without any comments. It might be useful to comment on the translators.

Jorge de Sena (1919-1978) translated Emily Dickinson. He was a Portuguese-born poet and university lecturer who lived exiled for six years in Brazil and, after the coup in 1964, immigrated to the United States where he worked as an academic. The fact that he had Portuguese roots was taken into account when comparing his translation with Google’s.

Geir Campos (1924-1999) translated Walt Whitman. He was a Brazilian poet and university lecturer. He lived his whole life in Brazil. He studied French, English, German, Italian, Spanish, Russian and Greek. He translated a total of thirty texts from all the languages he learned, including nine poetry titles, six novels and four plays.3 3 https://dicionariodetradutores.ufsc.br/pt/GeirCampos.htm

Pedro Gonzaga (1977) translated Charles Bukowski. He is a poet, translator, musician, writer and broadcaster. He translated seven books by Charles Bukowski.4 4 https://dicionariodetradutores.ufsc.br/pt/PedroGonzaga.htm

The translations

A Letter is a Joy of Earth

Jorge de Sena translated Emily Dickinson’s “A Letter is a Joy of Earth” probably in Araraquara in 1962 (DickinsonDickinson, E. (1979). 80 Poemas. (J. Sena de, Vert.). Edições 70., 1979). It was published posthumously in 1979 (for more information on Sena translating Dickinson see Monteiro, 1982Monteiro, G. (1982). Jorge de Sena’s Dickinson. Luso-Brazilian Review, 19(1), 23-29.).

Sena obviously understands the poem, whereas Google does not, and cannot relate the words to their context other than in terms of form. When Google succeeds in reproducing an alliteration or an assonance this is pure coincidence, although this coincidence occurs more than one would expect. An example is “At Deity decree”, translated by Sena as “Conforme Deus estatui”, Google: “No decreto da Deidade”. Google reproduced the assonance (d/d), not present in Sena’s translation. Besides, Sena translates 'Deity' as 'Deus', which is debatable, whereas Google correctly translated 'Deity' by ‘Deidade’.

Google blunders much less than one would expect. This might in fact indicate that Dickinson’s poetic language is more predictable than expected. Maybe for the same reason, Google’s translation is sometimes better than Sena’s, since the Portuguese author has a tendency to elevate the rather plain register of Dickinson’s poetry, albeit alternated with outdated language forms. The second verse of the poem is a case in point: “It is denied the Gods”, translated by Sena as “Denegada aos Deuses” and plainly rendered by Google as “É negado aos deuses”. Google respects the informal level of the original. Sena is more formal, but he reproduces the alliteration. Sometimes Sena privileges a poetic devise, missing an essential content feature such as in: “And I'm a Rose!”, which Sena translates as “E sou uma Rosa!”, whereas Google translates: “E eu sou uma rosa!”. Google and Sena both drop the rhyme (Breeze/trees), but the Google translation is more literal and seems more correct. Sena omitted the 'I', privileging the rhythm but the ‘I’ is essential. Sometimes neither of the translations entirely convinces. In “There's something in the flight”, Sena elevates the register, as he usual: “Algo há na fuga silente”. “Silente” is a very unusual word, also in Portugal, and furthermore there is no adjective in the original. Google, on the other hand, interprets ‘flight’ in its most literal, modern and common version: “Há algo no vôo”.

In general, Sena’s translation is the result of quite some analysing and reflection. This shows in his observance of the traditional poetical elements of the original (the complete comparison can be consulted in the Appendix). On the other hand Sena, being a human being, “interprets”. This is sometimes beneficial, as when elements have to be put in a relationship, especially when these elements are rather far apart. Sometimes, however, a more literal translation would have been truer to the original. As previously mentioned, Sena also has a tendency to elevate the general register of the poem. Dickinson occasionally uses outdated features, but this does not mean the whole poem is as solemn as Sena seems to insinuate.

To a Stranger

Geir Campos’ translation of Walt Whitman was published in 1964Whitman, W. (1964). Folhas de Relva. (G. Campos, Vert.). Rio de Janeiro: Ed. Civilização Brasileira. (Whitman, 1964). Quite a few of the comments to be made on Jorge de Sena’s translation of Emily Dickinson apply to this translation as well. Geir Campos very often “ennobles” the original in his translation, using words and turns of phrase pertaining to a register more elevated than the original. The title is a case in point: Campos translated “To a Stranger” as “A um Ser Estranho”, whereas Google translated plainly as “Para um estranho”. “Stranger” is a polysemous word. It can be a person one doesn’t know, or someone who does not belong to a particular place. In Portuguese, however, the most obvious translation, “estranho”, is equally polysemous, with the same sense possibilities. There is no obvious need to translate as “ser estranho”. Geir Campos obviously “explained” the original and gave it a somewhat more mysterious touch. The second verse was translated as follows:

Dickinson Campos Google Passing stranger! you do not know how longingly I look upon you, Estranho ser que passas! não sabes com que ansiedade ponho meus olhos em ti, Passando estranho! você não sabe o quanto ansioso eu olho para você,

Here the inevitable drawbacks of machine translation catch the eye. Google translated “Passing” as a Present Participle and “how longingly” by “o quanto ansioso”. The first translation is an outright mistake, the second one might sound a bit odd.

In the next verse, however, one can doubt if Google did not translate more accurately.

You must be he I was seeking, or she I was seeking, (it comes to me as of a dream,) bem podes ser aquele que eu andava buscando ou aquela que eu andava buscando (isso me ocorre como num sonho), Você deve ser ele que eu estava procurando, ou ela estava procurando, (vem a mim como de um sonho,)

Obviously “ele” as a translation of “he” is incorrect (aquele), but “You must be” is, in my opinion, more accurately rendered by “você deve ser” than by “bem podes ser”. Apart from this, the translation by Brazilian Geir Campos sounds rather peninsular (podes, buscar), which in Brazil means elevating the register. This is also the case in the next verse, where “somewhere” is translated by “algures” (“em algum lugar” by Google). There are a number of other examples of this characteristic (afeiçoados, cresceste, fôste, tomas-me a barba) Besides this, Google sometimes simply knows more English than the translator as in the expression “I am to + verb”, which Campos translates as follows:

I am not to speak to you, I am to think of you when I sit alone or wake at night alone, eu não estou para falar contigo, mas para pensar em ti quando me sento sozinho ou quando à noite desperto sozinho, Não devo falar com você, devo pensar em você quando eu me sentar sozinho ou acordar sozinho na noite, I am to wait, I do not doubt I am to meet you again, estou à espera, não duvido de que estou para encontrar-te outra vez, Eu espero, não duvido que eu seja para te conhecer de novo, I am to see to it that I do not lose you. com isso estou por ver que não te perco. Preciso assegurar que não o perca.

Surprisingly Google does not Translate all these sequences in the same way (Não devo, devo, Eu espero, preciso) and, unsurprisingly, errs once. Campos, on the other hand, translates all occurrences the same way, not correctly, as far as I can judge.

A number of other observations can be made (all to be consulted in the Appendix), but all come down to the same: Google blunders every so often, but Campos also isn’t without fault. Besides, Campos has a “plan” and his translation is coherent according to this “plan”. He knows who or what he is referring to when he speaks about a “Ser Estranho” and this fact conditions the rest of the translation. For whoever prefers a more solemn tone for this poem, Geir Campos is a “safe bet”. Those who prefer a more down to earth version might not be totally disappointed by the Google version.

Sandra

“Sandra” by Charles Bukowski was translated by Pedro Gonzaga. One of the most striking and also most expected features of the Google translation is again the machine’s difficulty in choosing the right alternative in the case of polysemous words. Polysemy is a troublesome problem for a machine, since it will immediately choose the most common alternative. In “Sandra” there are several examples of this: “as” translated as “como” instead of “enquanto”, “light” translated as “luz” instead of “acender”.

as she attempts to enquanto ela tenta como ela tenta Google fails in the translation of the homonym “as”. light acender luz Google is wrong in the translation of the homonym “light”. a new cigarette on an um novo cigarro num um novo cigarro em um

Google also has no way to always distinguish between “ser” and “estar”, although the context helps in some cases. The “literality” of Google often induces mistakes as in “spirit”, which Google translates as “espírito”, instead of “alcool” but Gonzaga as well errs.

she’s always high está sempre alta ela é sempre alta Google doesn't distinguish between “ser” and “estar”. The ambiguity of “high” is lost. in heels em sapatos de salto nos saltos Gonzaga explains. The Google translation is not quite understandable. spirit espírito espírito Gonzaga and Google have translated literally, but this is most likely about “alcohol”. pills boletas pílulas Gonzaga curiously translates this as “boletas”. It's probably drugs Bukowski is talking about. booze trago bebidas alcoólicas Gonzaga captured the informal register. Google formalized.

Google also does not always respect the collocations and translates break a man’s heart as “partir o coração de um homem”. “Quebrar o coração de um homem” is not so common (I leave the judgment of the choice of both of “de um homem” to my reader). On the whole, however, and when compared to Google, Gonzaga elevates the register, just like Sena and Campos.

Conclusion

The most obvious conclusion is that the machine gets it right more often than expected. The translations by Google are readable and would probably have been even more so if the verses had been joined together to make sentences and then broken up again. The astonishing fact is that a programme that relies basically on statistics and the analysis of the most likely sequence of words gets it right so often. Especially in the case of poetry where words are supposed to be chosen almost one by one, instead of in chunks, thus reducing predictability.

Google is also able to correct a wrong interpretation by the human translator and can, at least, function as a kind of safeguard for the human translator. In every one of the three poems analysed, Google had it right at least once where the human translator missed. Google errs where one expects it to err: polysemous words, gender, number and, in general, every time there is no distinction needed in the source text where there is a mandatory one in the target text. As Roman Jakobson is said to have said: “Languages differ less in what they can express than in what they must express”. This is especially significant in Machine Translation.

Besides, the comparison of the machine-translated poems with the translator-translated ones allows for a zero degree comparison that reveals the idiosyncrasies of poetry and shows that the computer effectively has problems with what are considered to be these characteristics of poetry: alliteration, assonance, rhythm and polysemy. Only a programme designed specifically to detect these features would be able to spot these. Having said that, the human translator mostly struggles with the same problems as the machine. My analysis shows the various layers present in poetry and how, between these different layers, translators have to choose the ones they will translate (alliteration, assonance, rhythm, repetition), unable to translate all of them all the time. According to my analyses, it is very rare for a translator to be able to render each and every one of these features every time.

The comparison of the computer-translated poems with the translator ones also provides a zero degree of comparison revealing the translators’ idiosyncrasies. In this case, maybe fortuitously, all three translators tended to use a more elevated register when compared to the plain Google translation. In some cases, the Google translation seemed to be more justified and closer to the original.

One caveat. This research is a snapshot. Technology evolves every day and the Google translation done in 2017 will now possibly be very different. Besides, as of October 2018, DeepL, the new automatic translation programme which started its activities in August 2017, has included Portuguese as one of its working languages. Translations are constantly getting better, and this will have an influence, acknowledged by translators or not, on every kind of translation and, why not, poetry translation.

References

  • Besacier, L. (2014). Traduction automatisée d’une oeuvre littéraire: une étude pilote. Traitement Automatique du Langage Naturel (TALN). Consulted https://hal.inria.fr/hal-01003944/document
    » https://hal.inria.fr/hal-01003944/document
  • Dickinson, E. (1979). 80 Poemas. (J. Sena de, Vert.). Edições 70.
  • Ghazvininejad, M., Shi, X., Priyadarshi, J., & Knight, K. (2017). Hafez: an Interactive Poetry Generation System. Proceedings of ACL 2017, System Demonstrations, 43-48.
  • Jiyuan Zhang, Yang Feng, Dong Wang, Yang Wang, Andrew Abel, Shiyue Zhang, Andi Zhang. (2017). Flexible and Creative Chinese Poetry Generation Using Neural Memory. Consulted https://arxiv.org/abs/1705.03773
    » https://arxiv.org/abs/1705.03773
  • Monteiro, G. (1982). Jorge de Sena’s Dickinson. Luso-Brazilian Review, 19(1), 23-29.
  • Qixin Wang, Tianyi Luo, Dong Wang, Chao Xing. (2016). Chinese Song Iambics Generation with Neural Attention-based Model. Consulted https://arxiv.org/abs/1604.06274v2
    » https://arxiv.org/abs/1604.06274v2
  • Sinclair, J. (1995). Corpus, concordance, collocation (3. impr). Oxford: Oxford Univ. Press.
  • Toral, A., & Way, A. (2018). What Level of Quality Can Neural Machine Translation Attain on Literary Text? In J. Moorkens, S. Castilho, F. Gaspari, & S. Doherty (Red.), Translation Quality Assessment (Vol. 1, pp. 263-287). Cham: Springer International Publishing. https://doi.org/10.1007/978-3-319-91241-7_12
    » https://doi.org/10.1007/978-3-319-91241-7_12
  • Toral, A., Wieling, M., & Way, A. (2018). Post-editing Effort of a Novel With Statistical and Neural Machine Translation. Frontiers in Digital Humanities, 5. https://doi.org/10.3389/fdigh.2018.00009
    » https://doi.org/10.3389/fdigh.2018.00009
  • Whitman, W. (1964). Folhas de Relva. (G. Campos, Vert.). Rio de Janeiro: Ed. Civilização Brasileira.
  • 1
    It is interesting for literary scholars to take a look at the drawbacks cited by scientists when it comes to the translation of poetry by machines. For example, [xref ref-type="bibr" rid="r3"]Ghazvininejad, Shi, Priyadarshi & Knight (2017[/xref], p. 43) cite three disadvantages. Number 3 is the next one: "Slow generation speed. Generating a poem may require a heavy search procedure. For example, the system of Ghazvininejad et al. (2016) needs 20 seconds for a four-line poem. Such slow speed is a serious bottleneck for a smooth user experience, and prevents the large-scale collection of feedback for system tuning. Certainly not a consideration that a literature scholar would make.”
  • 2
    "Il reste bien sûr des imperfections, des lourdeurs, voire des erreurs ponctuelles, qui appellent une correction" Principales erreurs ? - "Le défaut le plus répétitif, celui dont souffre d’ailleurs le travail de tout traducteur débutant, est le calque syntaxique, là où le français structure différemment la phrase .../... On comprend, mais ça ne sonne pas vraiment français" - "Autre défaut assez fréquent, la perte des idiomatismes du français au profit d’anglicismes. (…) " - "Un troisième défaut tient à la non prise en compte de certains repères culturels .../... Par exemple, Powers fait plusieurs références à la topographie de Boston qui donnent lieu à des inexactitudes dans la traduction : « la rivière Charles » par exemple (p. 12) qui n’est pas une rivière mais plutôt un fleuve ; c’est pourquoi on traduira par « la Charles River » ou simplement « la Charles » (…)"
  • 3
    https://dicionariodetradutores.ufsc.br/pt/GeirCampos.htm
  • 4
    https://dicionariodetradutores.ufsc.br/pt/PedroGonzaga.htm

Appendix. Comments on the translations by the translators and Google

Emily Dickinson
A Letter is a Joy of Earth

Walt Whitman
To a Stranger

Charles Bukowski
Sandra

Publication Dates

  • Publication in this collection
    29 Aug 2019
  • Date of issue
    2019

History

  • Received
    11 Jan 2019
  • Accepted
    13 Mar 2019
Universidade Federal de Santa Catarina Universidade Federal de Santa Catarina, Centro de Comunicação e Expressão, Bloco B- 405, CEP: 88040-900, Florianópolis, SC, Brasil, Tel.: (48) 37219455 / (48) 3721-9819 - Florianópolis - SC - Brazil
E-mail: ilha@cce.ufsc.br