Navigation – Plan du site

AccueilNuméros38PerspectivesMixing human and nonhuman dance p...

Perspectives

Mixing human and nonhuman dance partners: contrapuntal translation and the shift from replacement to inclusivity

Samuel Trainor

Résumés

Cet article remet en question la pertinence du paradigme sociologique du remplacement technologique en ce qui concerne la TAN dans le domaine littéraire. Latour (1988) appelle ce processus une « traduction ». Sa métaphore provient d’une conception erronée qui réduirait la traduction à une espèce d’instrumentalisation. Les débats autour de la TAN sont souvent hantés par ce spectre socio-économique. Cela n’a rien d’étonnant. La plupart des plates-formes TAN sont implicitement conçues pour « éliminer l’intermédiaire ». D’ailleurs, elles visent toujours la « domestication ». Leurs concepteurs voudraient qu’elles réussissent un « test de Turing » : que le lecteur lambda ne distingue pas leurs simulacres du véritable texte produit par un locuteur natif. Il s’agit ici du modèle classique de l’« invisibilité » du traducteur, modèle idéologique qui prédomine dans les méthodologies industrielles d’évaluation de la traduction. Il suffit d’un bagage traductologique de base pour y discerner la prémisse fausse. S’agissant de la littérature, les enjeux sont particulièrement élevés. Au sein de l’industrie technologique, la traduction littéraire est souvent vue comme un obstacle dans la quête pour l’IA générale. Cependant, les professionnels de la traduction littéraire souhaitant prévenir la TAN devraient éviter des objections théoriques absconses. Mieux vaut favoriser la promotion d’une éthique inclusive en traduction professionnelle. Si la TAN pourrait être adaptée à un fonctionnement polyphonique auprès des « bio-traducteurs », tout au long du processus (inclusion), plutôt que de reléguer l’intervention humaine à la « post-édition » (gage supposée de l’accessibilité d’un produit), il se peut qu’elle devienne un outil véritablement valable en traduction littéraire.

Haut de page

Texte intégral

  • 1 Serialised in the Red Republican from 9 November 1850.
  • 2 Engels Letter to Friedrich Adolph Sorge, 17 March 1872 (Marx and Engels, 1982: 342-344; letter 20 (...)
  • 3 Moore’s translation was published in 1888, three years after Lafargue’s, suggesting that the choi (...)

1The first English translation of the Communist Party Manifesto was published in 1850 by the Scottish feminist Helen Macfarlane.1 At the time, this seminal document was a curiosity. Much of the political and economic jargon we now take for granted (like “Proletariat”) may have seemed grotesque. So Macfarlane can be forgiven for amplifying the gothic rhetoric in the first sentence: “A frightful hobgoblin stalks throughout Europe.” (Marx and Engels, 1850: 1) Engels considered this to be almost as frightful a hobgoblin of a translation as Bakunin’s Russian version, published in Geneva in 1869.2 In the wake of the Paris Commune, in 1871, Marx and Engels’ obscure pamphlet enjoyed a revival on the international stage, due in no small part to a reissue of Macfarlane’s version in a New York feminist periodical, Woodhull and Claflin’s Weekly The first published French version, which appeared the following year in Le Socialiste (New York), was an anonymous retranslation of Macfarlane’s version. The French translator stumbled on the word hobgoblin. The first sentence goes down in a blaze of mistranslation: “Une terrible conflagration menace l’Europe.” (Marx and Engels,1871: 3) Later on, however, where Macfarlane has “the bugbear of Communism” (Marx and Engels, 1850: 1), the French is much more apt: “l’épouvantail du Communisme” (Marx and Engels,1871: 3). In response to this transatlantic phantasmagoria, Engels—always at pains to control the narrative—marshalled a new wave of retranslations, thereby retrospectively fulfilling the Manifesto’s claim to have been simultaneously published in a broad spectrum of European languages. Thus was Macfarlane’s foreignizing hobgoblin hunted down, along with its malformed French progeny, and replaced with the spectre translations of the 1880s, penned by Laura (Marx) Lafargue and Samuel Moore.3

  • 4 Engels makes this explicit in his letter to Sorge: “the French version especially is quite indisp (...)

2It is no coincidence that Moore should have mirrored Lafargue’s choice of spectre for the text’s first noun. The lexical alignment is emblematic of the process of standardisation imposed by Engels in his effort to maintain the appearance of a united front. So the word, in this context, has arguably become a symbol of suppressed linguistic diversity, the eradication of a cultural plurality that was seen to reflect the anarchism of Bakunin and Proudhon.4 It is emblematic of a revisionist denial of stylistic counterpoint and interpretative freedom in the document’s retranslation. The hobgoblin was sacrificed to the spectre of consistency.

  • 5 Ralph Waldo Emerson, “Self-reliance”, Essays, Cambridge, MA, Houghton & Mifflin, 1876, p. 52.
  • 6 Marx-Engels’ term is one of the German words for a ghost. Unlike the more common cognate Geist, i (...)
  • 7 Results of a comparative study of lexical collocations in political journalism for the year 1848 (...)

3However, “a foolish consistency,” wrote Ralph Waldo Emerson in 1841, “is the hobgoblin of little minds”.5 Macfarlane’s hobgoblin was not as strange in 1850 as it might appear in retrospect, nor is it incorrect. Rather than an eidolon, the original text, composed against the backdrop of the Campagne des banquets in France and a rebellion brewing in Vienna, evokes a bogey man. “Communism”, it says, has become a scare-mongering slur. Gespenst6 is just a metonym. It is not at all presented as a vision from beyond the grave. In 1848, when she began working on her translation, the words hobgoblin and bugbear, used by Macfarlane, were just as current in British journalism about revolutionary politics as were their French equivalents, croquemitaine and épouvantail7 in the French press. They were used as often as the word spectre in formulae that precisely mirrored Marx and Engels’ turn of phrase. So Macfarlane appears to have used apt contemporary vocabulary to capture the tongue in cheek histrionics of the German Manifesto’s curtain-raiser.

4A confession: the preceding paragraphs were first drafted in the 1990s. Their historical content may seem tangential to the question of contemporary machine translation (MT) in literature, but it is hoped the reader will forgive this disinterment of an old research project if a few key points of relevance can be enumerated. The first is a matter of historical coincidence. When this analysis of the translation history of the Communist Manifesto’s incipit was begun, Systran had just launched the world’s first online MT tool, Babel Fish. Two major books had also just been published that were to prove seminal in their fields, establishing concepts that remain applicable to the problematic effects of MT technologies on the practice of literary translation. The first was Derrida’s Spectres de Marx (1994), which extrapolated Lafargue’s sentence into an exposition of his evocative concept of hauntology. The second was Lawrence Venuti’s The Translator’s Invisibility (1995), which established the importance of examining the history of politically inflected strategies of translation. The application of the lessons of the latter to the former was the goal of the original analysis, but it was just the overture to an early experiment with machine translation of a literary text. This basically involved feeding extracts of the Communist Manifesto into Babel Fish. The results of this ad hoc investigation were to be analysed in light of the theoretical nexus of Derrida and Venuti. This is the hinterland of the current study, the intention of which is to take a diachronic critical view, rather than to offer a synchronic analysis of machine translations of a given literary corpus. This article is therefore something of an argumentative excursus. It seeks to advocate a greater role for translation theorists in developing both automatic translation technologies and new uses of AI in the creative practices of literary translation.

5Secondly, it is undeniable that the specific political ideas in the Manifesto—the Marxian concepts of technologically inflected historical materialism and labour-value—remain entirely relevant to debates surrounding the role of AI in professional translating. This is not just a matter of Luddite tub-thumping. If AI language composition and NMT have come on leaps and bounds over the past twenty years, it is arguably not the result of any major breakthroughs in the academic discipline of artificial intelligence, but attributable to an exponential rise in computing power allied to the systematic accumulation, by corporate technology giants, of vast parallel datasets (Bridle, 2023). The value of these corpora derives from the unacknowledged and uncompensated work of generations of human translators, writers, and internet users, including those prompting and correcting the output of the AI tools themselves. The production of human language has thus been industrialized and globalized. This is just one part of what is arguably the most extensive appropriation of labour-value in history: the mass collection and instrumentalization of online content, including historical archives of humanity’s creative output.

  • 8 These “image generators” are essentially designed to perform intersemiotic NMT.

6Finally, Macfarlane’s gothic term, a frightful hobgoblin is a rhetorical device whose salience in its historical context finds echoes in the era of AI. A similarly baleful phantasmagorical figure stalks contemporary discourses of digital cultures, and by extension those of Translation Studies. Rather than present this problem through the Derridean prism of hauntology, by calling AI a spectre—as the editorial from Spheres did in 2019 (Ganesh & Lohmüller, 2019) —we might prefer to emulate Macfarlane’s more modish style by using a word that is trending in contemporary AI journalism: the crungus. This is the name of a bogey man figure, created in response to a single (nonsense) word prompt by the OpenAI image generator tool, Craiyon.8 It is a visual signified generated by an AI from an empty verbal signifier. The crungus went viral in 2022. With this new bogeyman lurking in the shadows, perhaps we can summarise this third point by revisiting the Communist Manifesto’s incipit: a crungus has gone viral in the world of translation: the crungus of AI.

  • 9 While the general economic consensus has always suggested the threat of technological unemploymen (...)

7The broader fear of replacement by AI is nothing new. In 1863, pre-empting Marx’s post-Darwinian analysis of technological evolution in Das Kapital, Samuel Butler wrote of machines, “we are daily giving them greater power and supplying by all sorts of ingenious contrivances that self-regulating, self-acting power which will be to them what intellect has been to the human race. In the course of ages we shall find ourselves the inferior race” (Butler, 1914 [1863]: 182). However, it is the thesis of this article that the cultural fear of substitution by AI—embodied as the lurking crungus—whether or not the so-called “luddite fallacy” has any economic validity,9 derives from a fundamental humanist paranoia of technical substitution which is deeply ingrained in industrial capitalism (Simondon, 2014), and which finds its most comprehensive systemic analysis in the French tradition of la sociologie de la traduction. It turns out that literary translation theorists are uniquely well equipped to contest the basic terms of this idea.

Before discussing how this might be done, we need, in line with our diachronic approach, to go back to that translatological coincidence in the nineties: the simultaneous appearance of the first online machine translation platform and of Venuti’s seminal “History of Translation”. The impact of Babel Fish was crucial. Systran’s software was an automatic translation platform, which applied rules-based algorithms to synthetic language datasets, rather than a modern neural net trained with genuine corpora. At the time, in the context of Venuti’s politicized concept of translation strategy, the unprecedented possibilities for satirical experimentation in translation criticism opened up by the new MT platform seemed too good to pass up. Initial experiments with the Communist Manifesto were promising, but ultimately less informative than the original historical variations outlined above. They did, however, show that transfers between German and English turned out some strange results, presumably because Systran’s buggy algorithms were as yet unable to handle the differences of syntax and morphology. So, under the influence of the bug metaphor, the focus of the study turned to another seminal incipit in German literature, the beginning of Kafka’s “Die Verwandlung”. The results were uncanny. Below is the output produced by Babel Fish in 1998:

  • 10 The German original (Kafka, 1917 [1915]): Als Gregor Samsa eines Morgens aus unruhigen Träumen er (...)

As Gregor Samsa of one morning from jerky dreams awaked, he was changed in his bed to a tremendous vermin. It was situated on its tank-like hardback and saw, if it lifted the head a little, its curved, brown antinode bulge, on whose height the blanket, divided by arc-shaped reinforcements, to the complete ready, could keep hardly still. Its many compared with its other scope, pitifully thin legs flickered it helplessly before the eyes. (BabelFish.com, 1998)10

It is impossible to reproduce this output with the tools currently available. A legacy version of Systran’s Babel Fish can be found online, with the original URL (www.babelfish.com). However, it reflects the incremental improvements implemented since 1995. Many of the bugs have been exterminated. So the text above is a historical artefact. This synchronic contingency is a structural problem for academic research in a field in which technological advances far outpace the peer-reviewed literature, and the industry—intent on moving fast and breaking things—shows little interest in facilitating diachronic investigations. This has tended to force contemporary translation scientists to focus on providing snapshots of the state of play. Duncan Large, for example, has published a diverting comparison of various recent machine translations of Kafka’s incipit (Large, 2020). Regrettably, he only tested the first sentence, which the above quotation reveals to be less interesting than the rest of the paragraph.

8In 1998, the accidental surrealism of the Babel Fish versions of Kafka was compelling. The lexical oddities proved thought-provoking. Netherglide, a morphological calque of the German compound lexeme Niedergleiten, sounds like the name of an exotic creative space, a portmanteau netherworld in which everything floats away: Alice through the netherglide. And when the results were analysed in terms of textual cohesion (Halliday and Hasan, 1976), a critical issue was revealed that no published English translation has ever explored.

9From the second sentence onwards, Babel Fish used the neuter pronoun it to refer to the central character. The critical serendipity of this was obvious. It was the embodiment of posthuman translation. It was no less compelling for being obviously the result of some kind of bug. This word is worth emphasising not merely to hammer home the pun, but also to recall its earliest denotation in English: “An imaginary evil spirit or creature; a bogeyman” (OED 3rd Ed., bug, n.1): precisely the kind of thing Marx and Engels had meant by Gespenst. Glitches have always made humans think there might be a ghost in the machine. The point being that Babel Fish’s uncannily meaningful use of it can only be attributable to a technical glitch. The pronoun used in German is the masculine er, rather than the neuter es, which Kafka seems to have eschewed. In traditional translation practice, this would preclude the pronoun it, because it flaunts an implicit authorial choice. However, this kind of deviation is precisely the kind of syncopated creativity that my own translation theory seeks to encourage, especially in the retranslation of canonical texts. More on this below.

10If we compare Babel Fish’s buggy oddity to the first English translation by Willa Muir, the contrast is stark.

As Gregor Samsa awoke one morning from uneasy dreams he found himself transformed in his bed into a gigantic insect. He was lying on his hard, as it were armour-plated, back and when he lifted his head a little he could see his domelike brown belly divided into stiff arched segments on top of which the bed quilt could hardly keep in position and was about to slide off completely. His numerous legs, which were pitifully thin compared to the rest of his bulk, waved helplessly before his eyes. (Kafka, 1971 [1933]: 89)

Gregor Samsa could not sound more pathetically human. Muir’s remains the most elegant English version, despite its dated lexis, but its elegance is achieved via a domesticating approach. This was characteristic of translation strategies in the period and has inevitably been offset by some of the more recent versions. A key issue has been the lexical difficulty of rendering the German Ungeziefer, which names the creature into which Gregor Samsa is transformed. Where Muir gives us insect, Stanley Corngold’s 1972 version has vermin, a highly foreignizing choice (especially as the word is rarely used as a count-noun), adopted by several subsequent academic translators, like Joachim Neugroschel (1993) and Joyce Crick (2009). On the other hand, more creatively inclined versions, like those of Michael Hofmann (2007) and Christopher Moncrieff (2014), have taken the opposite approach, extending Muir’s hyponymisation with, respectively, cockroach and bedbug. It is worth noting the contrapuntal freedom taken here, in stark contrast to the reluctance among the French retranslators of the Communist Manifesto to deviate from the (authorized) word spectre. In any event, the bug that interests us is technical, rather than insectoid. None of these versions—not even Neugroschel’s foreignizing effort—approaches anything like the level of defamiliarization achieved by the buggy Babel Fish output in 1998.

11Today, of course, it is possible to compare NMT outputs with both the Systran one and these human versions, and we can see that the anthropomorphic transformation of the machine’s prose has altered it beyond all recognition.

When Gregor Samsa awoke one morning from restless dreams, he found himself transformed into a monstrous vermin in his bed. He lay on his armour-like hard back and, lifting his head a little, saw his bulging, brown belly, divided by arched stiffeners, at the height of which the bedspread, ready to slide down completely, could barely maintain itself. His many legs, pitifully thin compared to his other girth, flickered helplessly before his eyes. (DeepL [Free], September 2022)

There are a few lingering problems of semantic and cultural correspondence here (e.g. arched stiffeners), and the calqued syntax remains a little awkward, but these are minor differences when compared to the outlandish Systran output from 1998. Viewed alongside the recent human translations, however, especially those that reject the word vermin, it still comes across as clunky and uncreative. There seems little prospect of an NMT tool producing as witty a solution as Moncrieff’s creative synthesis, “changed into an enormous bedbug” (Kafka, 2014: 40), to translate “in seinem Bett zu einem ungeheueren Ungeziefer verwandelt”. Still, the ability of DeepL to output polished and almost plausible English prose can be impressive.

  • 11 See Wang et al., 2022 for a synopsis.
  • 12 See, for example, Baker, 1993; Laviosa, 2002; Kruger et al., 2013; Loock, 2016.
  • 13 See, for example, Koehn, 2009; Callison-Burch et al., 2011; Koponen, 2012: House, 2014; Toral and (...)

12So what has happened in the interim to explain this metamorphosis? It would take a more knowledgeable commentator to attempt an analysis of the specific technological advances,11 but, beyond the vast improvements in processing power and data handling, the two main differences between the Systran tool and DeepL appear to be: a) the switch from an abstract rules-based approach to a corpus-based methodology—mirrored by an analogous shift in linguistics-oriented translation science;12 and b) the application to a language processing tool of neural network technologies, which lend machines the ability to “learn” in ways that are loosely modelled on organic brain functions. What we can say with some certainty, however, is that the relevant literature shows MT developments to have been framed predominantly as a process of increasing the autonomy of output quality—i.e. minimising the apparent need for human editing—using various iterations of a heuristic model of translation quality assessment (TQA).13 The concept of autonomy is key, and its influence has skewed the methodology. There has been precious little consultation of translation scientists. Stephen Doherty acknowledges that:

while the field of MT has adopted TQA as a central aspect in its development, theoretical discussions of translation and translation quality are largely unknown in this field. Research in MT—like industry practice—instead frames TQA as part of a larger process, one in which it is a means to an end and not an end in itself. The pragmatism this entails, in contrast to the more principled approach sought after in TS, is not without its own shortcomings. (Doherty, 2017: 134)

Crucially, even where so-called “multidimensional quality metrics” (MQM) have been developed, the concept of quality applied has largely been couched in terms of the relative “acceptability” of a machine translation output (a “candidate translation”) as a viable replacement for a human translation output. This is what Reiβ and Vermeer would call a translatum (Reiβ and Vermeer, 2014 [1984], 1)—a single, empirical outcome of the complex social interactions involved in human translating. And the benchmark quality of the human translatum is itself routinely defined in terms of its own autonomous adequacy as a replacement for the source text. So MT outputs are usually product-tested, in artificial conditions, as replacements of replacements: the stand-in’s stand-in.

  • 14 Holz-Manttari’s theory of “Translatorial Action” (Schäffner, 2011) is broad enough to cover the p (...)

13As Doherty hints, appeals to pragmatism cannot justify the shortcomings of this approach. It is highly problematic in both theory and practice. It posits the possibility of language-switching without any meaningful intercultural hermeneutics and therefore precludes any translatological interrogation of methodology, which is ring-fenced as a technical consideration. The model of translatorial action14 propagated by the MT industry is therefore one in which the production of consumer-oriented linguistic substitutes is the sole function of a technical translation actor, and the role envisioned for the human professional is merely one of post-editing. My own contrapuntal theory of literary translation is diametrically opposed to this ideologically inflected model of TQA.

14The theory is posited on what I call the synoptic hypothesis: “New multimodal technologies of textual composition, edition, publication, broadcast and performance will make it increasingly common for multiple versions of texts to be received, modified and composed together.” (Trainor, 2017: 26) It is clearly not a Luddite doctrine. The theory’s most important tenet is a rejection of the primacy of what Christiane Nord calls “the instrumental function” of translation (Nord, 2018: 48-50): the assumption that a translation replaces (rather than accompanies) the original in the context of reception. The theory calls this the “instrumental replacement fallacy” and argues that the future of literary translation is collaborative, multimodal, interactive and inclusive. The basic terminology of translation studies is altered by the theory. Terms like baseline text and counterpoint are used, instead of source and target, and translations are treated as more like dance partners than substitutes. In practice, the theory analyses and propounds an array of interactive strategies of translation analogous to the interplay of musical polyphony. In particular, it identifies and encourages syncopated or off-beat translations of canonical literature, sometimes (though not necessarily) related to political reinterpretations, such as feminist retranslation (Trainor and Phillips, 2019).

  • 15 There are obviously exceptions. Michael Cronin would be one notable example of a translation scie (...)

15Seen from this theoretical standpoint, mainstream developments of NMT, informed by reductive models of TQA, appear to have three structural ideological flaws which remain resistant to the interventions of translation scientists. This is in part because we have too often acquiesced to the belief that such things are beyond the scope of our discipline.15

  • 16 See Steingard and Fitzgibbons, 1993. I am using TQM, inexpertly, as a loose umbrella term coverin (...)

16The first of these is the twentieth-century industrial ideology of productivist quality management, most notably Total Quality Management (TQM), which is linked to late-capitalist theories of supply side economics.16 This concept tends to subordinate all system-wide activities within an organisation, including those not directly engaged in production, to the maximization of product value. It does so primarily via a metaphor, conceptually reducing all useful activities to forms of production. The irony of the Marxian parallels here should not go unnoticed. Given the predominance of this kind of management ideology in the technical industries, including the industries with which translators tend to work, it is hardly surprising that NMT should be so affected by analogous flaws. However, these ideologies have had grave social and ecological consequences that go far beyond a reductive concept of translation. Many contemporary philosophers have highlighted the urgent need to break free of ideological productivism. Bruno Latour, for example, in one of his last published articles said, “il ne s’agit plus de reprendre ou d’infléchir un système de production, mais de sortir de la production comme principe unique de rapport au monde” (Latour, 2020). My translation theory falls squarely into this broader anti-productivist movement.

17The second structural problem in contemporary NMT developments is the subordination of the specific exercise of translation to a reductionist concept of general intelligence that is endemic in the broader field of AI research. There is a kind of déformation professionnelle at work. AI developers need to take an analytical approach to organic intelligence, dissecting its functions in order to translate them (as Latour would put it) into AI. As a result, the models they apply are often overtly consequentialist and treat efficiency of output as their overriding goal (as if inefficient behaviours like lingering, digression and complexification were antagonistic to intelligence—as if anything that reduced productivity also hampered intelligence).

  • 17 http://prize.hutter1.net/hfaq.htm [consulted 10 February 2022].

18The Hutter prize provides a good example. This award fund is a major driver in global efforts to produce Artificial General Intelligence (AGI). A large cash prize is offered for successful AI compression of natural language. Its goal is “to encourage development of intelligent compressors as a path to AGI.” It justifies this by saying that “compression ratios [in relation to natural language as a repository for human knowledge] can be regarded as intelligence measures”, even going so far as to define general intelligence as “an agent’s ability to achieve goals in a wide range of environments. The key concepts for a formal definition are compression and utility maximization, the other aspects are emergent phenomena”.17 On this evidence, key developers of AGI would appear to be as ignorant of epistemology as NMT developers are of translation science. The reductionist epistemological bias in NMT is therefore related to the fact that the overriding goal of AGI is often superordinate to the specific functions of specialised AI applications.

19Which brings us to the third ideological problem in current NMT development: the instrumental replacement fallacy—the notion that a translation can be judged solely in terms of its adequacy as a substitute for the original text. This last point is the crux of the argument. Bruno Latour provides us with a theoretical mechanism with which to interrogate the relationship between the replacement adequacy fallacy in translation-proper, which has been cogently refuted by generations of literary translation theorists, and the problem of what Latour and his colleagues call la sociologie de la traduction. This sociological model is particularly relevant as it concerns the delegation of human social functions to technological actors—like replacing a police officer with a set of traffic lights (Latour, 1988). This is precisely analogous to our own fear of human translators being supplanted by NMT. Translation professionals often seek to exorcise this crungus by heaping derision on its sillier outputs, especially when applied to canonical literary texts. Mark Polizzotti, for example, wraps up a chapter of his Translation manifesto with an anecdote about a Russian computer’s attempt to translate the English version of the Biblical axiom “The spirit is willing but the flesh is weak”: “With mechanical aplomb,” he scoffs, “it spat out what to its circuits must have seemed a perfectly sound analogue: ‘The vodka is strong but the meat is rotten.’” (Polizzotti, 2018: 47) And this kind of blooper-shaming is a common way of dispelling demons. But by engaging in this kind of thing, we acquiesce to the instrumental premise of substitution, which is the heart of the problem. It is striking how powerfully this concept has been allowed to dovetail in NMT development, aligning the age-old fantasy of translatorial invisibility with the concept of the Turing test. AI, like the transparent translation, seeks to pass unseen.

Most of the models of human quality assessment of machine translation outlined in the literature have focused on developing extended variations on the classic Turing test question: “can you tell this has been translated by a machine?” Which is also a variation on the classic instrumentalist translation question: “can you tell this has been translated?”, the assumption being that an affirmative answer to either question indicates relative failure. Surely, as translation scientists, having read Venuti and Berman, we should know better. Instead of playing this game, we should refuse to treat NMT outputs as potential substitutes for human translation of literary texts.

20The truth is that NMT engines—especially when they are applied to literary texts, in which an awareness of the problematic communicative interactions of cultures is often an important theme, are revealed to be nothing more than pseudo-translatum generators. The failures of NMT to handle metalingual and interlingual content in literature are often the best indicators of this shortcoming. For example, none of the available online NMT platforms I tested were able to cope with the rich depiction of intercultural communication issues in the following extract of A. S. Byatt’s short story “Crocodile Tears”:

[S]he heard a voice.
Excusez-moi, madame, est-ce que je peux vous aider ?She did not move. The tears ran.
He knelt at her feet, in his blue linen coat, clumsy on one knee. His French accent was clumsy, too. He changed to English, the excellent English of the Scandinavian North. […]
“Maybe I can help you to your room? Bring you a drink, perhaps. I cannot watch—I cannot watch—I should like to help. […] I know your room,” he said. “Come. We will go there, and then I will send for a drink”. (Byatt, 1998: 26-27)

Besides the predictable “effacement des superpositions de langues” created by the reductive choice of a single target language, every one of the NMT platforms into which this extract was fed failed to depict “the excellent English of the Scandinavian North” (Berman, 1985: 68) with anything approaching Byatt’s aplomb. Byatt captures the Norwegian’s English with great skill: it is grammatically excellent and perfectly easy to understand, but deictically strange, unidiomatic and lacking in natural orality. This would be a challenge for a human translator. But even an inexperienced professional would immediately recognise the cultural issues. The NMT platforms turned out a fumbling hotchpotch of the standardised, the unidiomatic and the downright incorrect. The DeepL example was one of the better attempts:

[E]lle entendit une voix.
Excusez-moi, madame, est-ce que je peux vous aider ? Elle ne bouge pas. Les larmes coulent.
Il s’est agenouillé à ses pieds, dans son manteau de lin bleu, maladroit sur un genou. Son accent français était maladroit, lui aussi. Il passa à l’anglais, l’excellent anglais du Nord scandinave. [...]
Je peux peut-être vous aider à monter dans votre chambre ? Vous apporter un verre, peut-être. Je ne peux pas regarder – je ne peux pas regarder – je voudrais vous aider. […] Je connais votre chambre, dit-il. Nous allons y aller, et ensuite j’enverrai chercher un verre. (DeepL [Free], September 2023)

All platforms provided the same calqued structure for “in his blue linen coat, clumsy on one knee”. On the other hand, there were wild variations for “Come. We will go there, and then I will send for a drink.” None of which rang true. The DeepL example above effaces the awkward deictic switch in “Come. We will go there”, giving only an overly unidiomatic version of the second element. Then it slightly misconstrues what is meant by “I will send for a drink”. It is, however, far superior to the attempt by Google Translate, which offered the unfathomable: “Viens. Nous y irons, puis j’enverrai boire un verre.”

  • 18 https://www.youtube.com/watch?v=fn3KWM1kuAw [consulted 5 January 2022]
  • 19 https://google.github.io/aichoreographer/ [consulted 20 March 2023]

21The point is that these kinds of examples make it clear how incapable NMT software is of genuine hermeneutic action. In short, these platforms do not actually translate at all. They leave no trace of interpretative intervention because there is none. If we perceive interpretations, these are merely the projections humans routinely perform to facilitate communication. It is analogous to ascribing motivations to the “interlocutor” when interacting with ChatGPT. NMT platforms are like the humanoid robots that mimic human dance moves, whether via programming (like the Boston Dynamics machines in the viral video “Do You Love Me?”)18 or training (like Google’s AI Choreographer platform, which “generates realistic smooth dance motion in 3D with full translation”).19 These technical constructs are not actually dancing. Such android technologies merely mimic the gestural end-result of human social activities. The better they get at this—the more they move or write like humans—, the more this shortcoming stands out. The DeepL version of Kafka’s “Die Verwandlung”, for all its polish, arguably has less value as a retranslation than the old Babel Fish version, whose inadequacy as a substitute highlights its value as a counterpoint.

22In the coming years, NMT will doubtless undergo further anthropomorphic improvements, with better targeting of corpora or output contexts. Perhaps this will include extended “training” of targeted literary AI platforms, beyond the usual method of feeding the neural network literary texts (Toral and Way, 2018). Increased collaboration with translation professionals is likely to help. On balance, this is probably to be encouraged, if only for sociological reasons, but in terms of translation methodology it will simply “infléchir un système de production”, as Latour puts it, rather than “sortir de la production comme principe unique de rapport au monde” (Latour, 2020). In its current form, NMT is an android technology, which is fundamentally constrained by its ambition to replace human activity via superstructural mimesis. And this problem is related to the crungus of AI replacement. Perhaps literature-focused Translation Studies, being the traditional domain of anti-instrumental theories, has something to teach the world in this regard: stop thinking in terms of replacement! When it comes to interactions between the technological and the human, Actor Network Theory (ANT) has already shown the way, building on Gilbert Simondon’s treatment of technologies as social actors.

  • 20 This is a bravura piece of literary social-science writing, which playfully foregrounds the probl (...)
  • 21 Michel Serres, Hermes III. La traduction, Paris, Les éditions de minuit, 1974. Latour and Serres (...)

23If we look back at one of ANT’s foundational texts, Latour’s witty early essay “Mixing Humans and Nonhumans Together: The Sociology of a Door-Closer” (Latour, 1988),20 it is clear that his account of sociological “translation” applied to human-nonhuman delegations is applicable to the relationship between human translators and NMT. Replacing a doorman with a mechanical door closer may be much simpler, but it is the same basic concept. However, Latour was a life-long opponent of productivism. His article emphasises the analysis of process and interaction within a paradigm of sociological translation, understanding technologies to occupy the same social space as humans, rather than acting merely as producers of output. In this sense, we could use the work of Latour to improve NMT developments, revealing NMT to have been designed thus far with scant regard for the social realities (the “scripts”) of translatorial action: i.e. with hardly any sociological “translation” at all. However, to do so, we need to drop the flawed metaphor of sociological translation. Latour’s essay aligns the idea with a teleology of human technological replacement. He is not wrong in his analysis of social processes involving nonhuman actors, but his metaphor of translation (no doubt gleaned from Michel Serres)21 is skewed by an application to human replacement, falling into the trap of assuming that translation is merely a process of functional substitution. The teleology is problematic, and this becomes obvious when applied to the processes of literary translation. As we have seen, contrapuntal translation theory treats the activity of translating literature and the literary translation object—a combination of the original text and its translata—as functionally inseparable: a multidimensional space of polyphonic interaction with no element of replacement or closure. It is therefore scarcely applicable to Latour’s account of how a doorman is replaced by a door closer. The same might be said of how a human translator of Kafka or A. S. Byatt might be replaced by NMT.

24So far, then, we have only identified the ideological flaws in current NMT developments. We now need to ask ourselves what can be done about them. The solution I am going to propose involves one final paradigm shift: an increasingly collaborative and inclusive approach to literary translation, and one that insists on the importance of neurodiversity as an ethical foundation. To do this, I propose an expansion of the term neurodiversity, in line with the fact that, very soon, AI hegemony and homogeneity is likely to render all human cognition that falls beyond its scope neuroatypical. This is not simply because we are becoming cognitively disabled in comparison to the vastly superior speed and capacity of AI. Nor is it because AI is intrinsically incapable of its own diversity. It is because the material reality and ideology of technological globalisation in which AI currently operates structurally negates diversity, favouring processes of data aggregation and allowing a single AI to do the work of a multitude of human actors, all of whom would have different mental and cultural particularities. Interculturality is erased by global NMT platforms to an extent that makes Engels’ standardisation of the Communist Manifesto translations look like small beer. If you were to travel to Japan and feed the first paragraph of “Die Verwandlung” into DeepL, you would get precisely the same output as a deaf student in North Africa, an Indian hijra, an old woman from the Amazon jungle, or an astronaut on the international space station. As Claire Larsonneur has pointed out, humans are already being systemically trained, via what Akrich’s sociology would call “la prescription” (Akrich, 2006), to mimic this homogeneity of language culture in order to work more productively with machines. So-called prompt-engineering happens in both directions. “Lack of diversity, standardization and control of language are,” she says, “iterative processes, so this trend may become massive.” (Larsonneur, 2019) In the future, any failure to adapt to this process of acute standardisation is likely to be perceived as deviation from a cognitive norm.

  • 22 In this regard, I have been influenced by research-actors working in inclusive theatre translatio (...)

25However, if we insist on inclusion as an ethical and creative imperative, in literary translation in particular, we might begin to break the cycle. Inclusion here does not refer to ensuring inclusive translation outputs, like gender-neutral texts. Well-intentioned as that kind of thing may be, it is just one more aspect of output-focused quality control. An AI could easily be trained to perform what is a fairly simple systematic procedure. Instead, inclusion is intended to refer to direct participation by diverse actors from the outset of any translation process. In disability studies, this is the conceptual distinction drawn between models of accessibility (focused on late-stage adaptation and consumer satisfaction), and models of inclusion (participation from the outset in design and implementation). We should stop, as it were, judging translation outcomes as accessible products and start promoting inclusive practices.22

26As an illustration of some of the key differences between accessibility and inclusion, in relation to contrapuntal creativity, the following is a transcript of the voice-over track in the inclusive dance film, Unspoken Spoken, released by the Candoco Dance Company in 2018. It is offered here both as a literary text in its own right, rather than merely an “audio-description”, and as an example of a kind of inclusive multimodal translating practice that integrates literary translation as a key aspect.

  • 23 Unspoken Spoken, directed and choreographed by Fin Walker, Candoco Dance Company, 2018. https://c (...)

The cookie-cutter shapes burst their banks. The floor turned into a river. The river into a dragon. The dragon spilt out of the classroom and lit up the sky. Whenever I watch people in groups, I’m reminded of that day. Any new idea – no, that’s not the word – any new normal, any new change begins with small, individual actions that multiply across the masses, until, in the end, no-one knows if it’s you who created the momentum, or whether the momentum created you.
From left to right, Dennis, Rosalia, Alice, Fred and Mavis are sat side by side in the black space from the opening scene. They look fed-up and are fidgeting and looking out to the viewer. Dennis and Rosalia stand up and sit down in unison. Alice joins in, mirroring their movements as they perch on the edge of their seats, as if to get up, before slumping back and coughing with a hand over the mouth. Gradually, Fred joins in, leaning forward and back, turning to look at Mavis, who is frowning. All the time the pace is increasing and the tension building. They all move forward to a huge table containing numerous vases of dead flowers and upturn the table. All the dead flowers and vases go crashing to the floor.
Back in the white space with the door in the background, for a moment there is an empty space, then Fred and Alice spin and glide around and over each other, wreathed in smiles.
23

Compared to the inclusive dance performance, the “audio-description” in this film clip is ostensibly an example of accessibility management, allowing visually impaired people consumer-access to the product. But that is a reductive reading. The text is much more complex and interesting than a mere description. Notably, it does not mention the visible disabilities of the dancers (Dennis has motor difficulties, Alice has no lower-limbs, and Mavis has upper-limb differences and uses a wheelchair). Instead it leans towards a more evocative and metaphorical style, with expressions like “they perch on the edge of their seats” and “wreathed in smiles”. Elsewhere it uses slightly defamiliarizing or even foreignizing expressions in references to bodily movement, saying, “a hand over the mouth” (a non-possessive form that sounds more like French than English).

27This is because the spoken soundtrack is not acting purely as a visual description. It is part of the artistic polyphony. Its stylistic mechanisms join in with the performance. In the first paragraph of the transcript, it incorporates an English translation of a highly literary text in British Sign Language: an important semiotic element of a deaf dancer’s performance, which shifts from a poetic narrative to a philosophical reflection, expressing doubt about its own lexical choices. The voice-over also interacts musically with the score and the dancers’ movements, in terms of rhythm, pace and tone. Sometimes the descriptions are tonally and rhythmically displaced. At other points, words, musical effects and movements coincide. It therefore exhibits literal aspects of musical counterpoint. Furthermore, it incorporates material that seems to come from pre-existing choreographic discourse, thus creating a simultaneous polyphony with elements of a diachronic process of artistic development. So while the audio description is ostensibly intended as an aid to client-side accessibility, when it is read in this way, as a multimodal contrapuntal translation performance—if it is read more like a literary translation—, it can be seen to have interactive elements that bring it into the functional polyphony of the inclusive choreography.

  • 24 https://teamyacht.com/projects/the-computer-accent [consulted 29 November 2022]

28Obviously, there is no AI element here. Recent examples do exist of artists and musicians incorporating AI as a creative element in polyphonic composition processes that might be called machine inclusive. The avant-garde American dance-pop group YACHT, for example, recently produced the album Chain Tripping (2022) using “cutting-edge data analysis tools, machine learning, neural networks, sci-fi instruments and generative composition strategies to create a new kind of human-machine music—melodies, lyrics, artwork, videos and all”.24 However, the Unspoken Spoken soundtrack is offered here more as an analogy than a direct example. If we can imagine translating as analogous to dance (an ongoing and expanding dance of cultures, texts, languages, ways of thinking and saying) then it is easy enough to imagine non-human dancers in the mix. But we would not want them to pass themselves off as human, or to replace human dancers, any more than we would want a dancer without physical or motor disabilities to dissimulate their relative agility in the presence of partners with different bodies and ranges of movement. This recalls the contrast between modernist synthetic electronic music (which has produced some truly amazing compositions and sounds over the years), and the more recent advances in mimetic sound modelling, which, while impressive in technical terms, are rarely put to any more creative use than to reproduce the complex tonalities of an acoustic instrument.

29In conclusion, then, if developers of NMT can be persuaded by literary translation theorists to resist the ideologically reductive, instrumentalist urge to prioritise the concept of replacement, and if translators can be encouraged to interact differently with AI technologies, as part of an inclusive, polyphonic process of interaction—boosted by the technological advances that are continually allowing texts to be more polyvalent, multi-modal, adaptive and adaptable, if diverse human actors can be encouraged to intervene from the outset in creative projects, and not merely in post-editing, then we could see literary translation evolve into something genuinely new. Crucially, this would also require an emancipation of AI from the reductive strictures of the replacement fallacy and surface-level anthropomorphism. Who knows what kinds of weird and wonderful metamorphoses an AI might reveal (and undergo) were it allowed to participate in the social processing of textual interactions between natural languages as a genuine actor: i.e. in ways that reflected its own (buggy) experiences of the world?

30Human and nonhuman translators of the world get down!

Haut de page

Bibliographie

Corpus Texts

Byatt, A. S., “Crocodile Tears”, Paris Review, Issue 146, Spring 1998, p. 1-41.

Kafka, Franz, Die Verwandlung, Leipzig, Kurt Wolff Verlag, 1917 [1915], https://www.gutenberg.org/ebooks/22367 [consulted 1 February 2022].

—, “The Metamorphosis”, trans. Willa and Edwin Muir, Nahum N. Glatzer (ed.), The Complete Stories, New York, Schocken Books, 1971 [1933], p. 89-139.

—, “The Metamorphosis”, trans. Stanley Corngold, New York, Norton, 1996 [1972].

—, The Metamorphosis, In the Penal Colony and Other Stories, trans. Joachim Neugrochel, New York, Simon & Schuster, 1993.

—, Metamorphosis and Other Stories, trans. Michael Hofmann, London, Penguin, 2007.

—, The Metamorphosis and Other Stories, trans. Joyce Crick, Oxford, OUP, 2009.

—, The Metamorphosis and Other Stories, trans. Christopher Moncrieff, Richmond, UK, Alma Classics, 2014.

Marx, Karl and Engels, Friedrich, Manifest der Kommunistischen Partei, London, 1848, Deutsches Textarchiv https://www.deutschestextarchiv.de/marx_manifestws_1848 [consulted 20 December 2021]

—, “German Communism—Manifesto of the German Communist Party”, transl. Helen Macfarlane, Red Republican, Vol. 1, No. 21, 9 November 1850, p. 1.

—, “German Communism—Manifesto of the German Communist Party”, transl. Helen Macfarlane, Woodhull & Claflin’s Weekly, 30 December 1871, p. 3-7, 12-13.

—, “Manifeste de Karl Marx”, (anonymous translation of Macfarlane’s English version), Le Socialiste, New York, No. 16, 20 January 1872, p. 4.

—, “Manifeste du parti communiste”, trans. Laura (Marx) Lafargue, Le Socialiste, Paris, No. 1, 29 August 1885, p. 4.

—, Manifesto of the Communist Party, trans. Samuel Moore, London, William Reeves, 1888.

—, Le Manifeste communiste, trans. Charles Andler, Paris, Bibliothèque socialiste, 1901.

—, Manifeste du parti communiste, trans. Francis Brière (1962), Paris, Bibliothèques 10/18, 1998.

—, “Le Manifeste communiste”, trans. Maximilien Rubel and Louis Évrard, Œuvres, Vol. 1, Économie I, Paris, Gallimard, 1963, p. 161-195.

—, Manifeste du parti communiste, trans. Émile Bottigelli, Paris, Aubier-Montaigne, 1971.

—, Manifeste du parti communiste, trans. Michèle Kiintz, Paris, Éditions sociales, 1972.

—, Manifeste du parti communiste, trans. Corinne Lyotard, Paris, Librairie générale française, 1973.

—, Le Manifeste du parti communiste, trans. Gérard Cornillet, Paris, Messidor/Éditions sociales, 1986.

Works Cited

Akrich, Madeleine, « La description des objets techniques », Sociologie de la traduction: Textes fondateurs [online], Paris, Presses des Mines, 2006, p. 159-178, https://books.openedition.org/pressesmines/1197 [consulted 10 April 2020]

Baker, Mona, “Corpus linguistics and translation studies: Implications and applications”, Mona Baker, Gill Francis, and Elena Tognini-Bonelli (eds.), Text and technology: In honour of John Sinclair, Amsterdam, Benjamins, p. 233-250.

Berman, Antoine, « La traduction et la lettre ou l’auberge du lointain », Les tours de Babel, Mauvezin, Trans-Europ-Repress, 1985, p. 35-150.

Bridle, James, New Dark Age. Technology and the End of the Future, London, Verso Books, 2023.

Butler, Samuel, (as “Cellarius”), “Darwin Among the Machines (To the Editor of The Press, Christchurch, New Zealand, 13 June, 1863)”, A First Year in Canterbury Settlement With Other Early Essays, London, A. C. Fifield, 1914, p. 180-185.

Callison-Burch, Chris, Koehn, Philipp, Monz, Christof et al., “Findings of the 2011 Workshop on Statistical Machine Translation”, Proceedings of the Second Workshop on Statistical Machine Translation, Prague, 2011, p. 136-158.

Derrida, Jacques, Specters of Marx. The State of the Debt, the Work of Mourning, and the New International, trans. Peggy Kamuf, New York, Routledge, 1994.

Di Giovanni, Elena, “Inclusive theatre-making: Participation, empowerment and well-being”, in TRAlinea, Special issue: Inclusive Theatre: Translation, Accessibility and Beyond, 2022, https://www.intralinea.org/specials/article/2609 [consulted 12 January 2023]

Doherty, Stephen, “Issues in human and automatic translation quality assessment”, Dorothy Kenny (ed.), Human Issues in Translation Technology, London, Routledge, 2017, p. 131-148.

Frey, Carl Benedikt and Osborne, Michael A., “The future of employment: How susceptible are jobs to computerisation?”, Technological Forecasting and Social Change, Vol. 114, January 2017, p. 254-280.

Ganesh, Maya Indira and Lohmüller, Stina, “Editorial #5 Spectres of AI”, Spheres, No. 5, November 2019, https://spheres-journal.org/contribution/5-spectres-of-ai/ [consulted 12 January 2023]

Halliday, M. A. K. and Hasan, Ruqaiya, Cohesion in English, London, Longman, 1976.

Han, Chao, “Translation quality assessment: a critical methodological review”, The Translator, Vol. 26, No. 3, 2020, p. 257-273.

House, Juliane, “Translation Quality Assessment: Past and Present”, Translation: A Multidisciplinary Approach, London, Palgrave Macmillan, 2014, p. 241-264.

Klubička, Filip, Toral, Antonio and Sánchez-Cartagena, Victor M., “Quantitative fine-grained human evaluation of machine translation systems: a case study on English to Croatian”, Machine Translation, Vol. 32, No. 3, September 2018, p. 195-215.

Koehn, Philipp, Statistical Machine Translation, Cambridge, CUP, 2009.

Koponen Maarit, “Comparing human perceptions of post-editing effort with post-editing operations”, Proceedings of the Seventh Workshop on Statistical Machine Translation, 7-8 June 2012, Montréal, 2012, p. 181-190.

Koponen, Maarit, Mossop, Brian, Robert, Isabelle S. et al. (eds.), Translation Revision and Post-editing, London, Routledge, 2020.

Kruger, Alet, Wallmach, Kim and Munday, Jeremy (eds.), Corpus-based translation studies: Research and applications, London, Continuum, 2013.

Large, Duncan, “Post-human Literary Translation? A Kafka(esque) Example”, Goethe Institut, 2020, https://www.goethe.de/ins/th/en/kul/mag/21984887.html [consulted 12 January, 2023]

Larsonneur, Claire, “The Disruptions of Neural Machine Translation”, Spheres, No. 5, November 2019, https://spheres-journal.org/wp-content/uploads/spheres-5_Larsonneur.pdf [consulted 12 January, 2023]

Latour, Bruno, (as “Jim Johnson”), “Mixing Humans with Non-Humans: Sociology of a Door-Closer”, Social Problems, Vol. 35, 1988, p. 298-310, http://www.bruno-latour.fr/sites/default/files/35-MIXING-H-ET-NH-GBpdf_0.pdf [consulted 10 April 2020]

—, « Imaginer les gestes-barrières contre le retour à la production d’avant-crise », AOC, 30 March 2020, https://aoc.media/opinion/2020/03/29/imaginer-les-gestes-barrieres-contre-le-retour-a-la-production-davant-crise/ [consulted 10 April 2020]

Läubli, Samuel, Sennrich, Rico and Volk, Martin, “Has Machine Translation Achieved Human Parity? A Case for Document-level Evaluation.” Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, Brussels, Association for Computational Linguistics, 2018, p. 4791-4796.

Laviosa, Sara, Corpus-based translation studies: Theory, findings, applications, Amsterdam, Rodopi, 2002.

Loock, Rudy, La traductologie de corpus, Villeneuve D’Ascq, Presses universitaires du Septentrion, 2016.

Moorkens, Joss, Castilho, Sheila, Gaspari, Federico et al. (eds.), Translation Quality Assessment, From Principles to Practice, Heidelberg, Springer, 2018.

Marx, Karl and Engels, Friedrich, Collected Works, Vol. 44, Letters 1870-73, trans. Rodney Livingstone, London, Lawrence and Wishart, 1982.

Nord, Christiane, Translating as Purposeful Activity: Functionalist Approaches Explained, 2nd Edition, London, Routledge, 2018.

Polizzotti, Mark, Sympathy for the traitor: a translation manifesto, Cambridge, MA, the MIT Press, 2018.

Popel, Martin, Tomkova, Marketa, Tomek, Jakob et al. “Transforming machine translation: a deep learning system reaches news translation quality comparable to human professionals”, Nature Communications, Vol. 11, Article no. 4381, September 2020, https://doi.org/10.1038/s41467-020-18073-9 [consulted 9 February, 2022].

Popović, Maja, “Informative Manual Evaluation of Machine Translation Output”, Proceedings of the 28th International Conference on Computational Linguistics, Barcelona, International Committee on Computational Linguistics, 2020, p. 5059-5069.

Reiβ, Katharina and Vermeer, Hans J., Towards a General Theory of Translational Action. Skopos Theory Explained, trans. Christiane Nord, London, Routledge, 2014 [1984].

Schäffner, Christina, “Theory of translatorial action”, Yves Gambier and Luc Van Doorslaer (eds.), Handbook of Translation Studies, Vol. 2, Amsterdam, Benjamins, 2011, p. 157-162.

Simondon, Gilbert, Sur la technique (1953 – 1983), Paris, PUF, 2014.

Steingard, David S. and Fitzgibbons, Dale E., “A Postmodern Deconstruction of Total Quality Management (TQM)”, Journal of Organizational Change Management, Vol. 6. No. 5, 1993, p. 27-42.

Toral, Antonio and Way, Andy, “Translating Literary Text between Related Languages using SMT”, Proceedings of the Fourth Workshop on Computational Linguistics for Literature, Denver, Colorado, Association for Computational Linguistics, 2015, p. 123-132.

—, “What Level of Quality can Neural Machine Translation Attain on Literary Text?”, Translation quality assessment: From principles to practice, 2018, p. 1-26.

Trainor, Samuel, “Cinema skopos: Strategic Layering and Kaleidoscopic Functionality in Screenplay Translation”, Palimpsestes, 30 | 2017, p. 15-46.

Trainor, Samuel and Phillips, Jess, “”From Chorus to Counterpoint: Carnivalesque Feminist Choreography in an Expanded Translation of Aristophanes’ Ecclesiazusae”, Coup de théâtre, 33, Traductions et adaptations des classiques dans le théâtre anglophone contemporain, 2019, p. 155-172.

Venuti, Lawrence, The Translator’s Invisibility. A History of Translation, London, Routledge, 1995.

Wang, Haifeng, Wu, Hua, He, Zhongjun et al., “Progress in Machine Translation”, Engineering, Vol. 18, November 2022, p. 143-153.

Haut de page

Notes

1 Serialised in the Red Republican from 9 November 1850.

2 Engels Letter to Friedrich Adolph Sorge, 17 March 1872 (Marx and Engels, 1982: 342-344; letter 209). Engels’ dismissal of this translation was almost certainly motivated by political differences with the anarchist Mikhail Bakunin, rather than any close knowledge of the content of the translation itself.

3 Moore’s translation was published in 1888, three years after Lafargue’s, suggesting that the choice of spectre can be attributed to Marx’s daughter. Subsequent published French translations include: Charles Andler (1901), Francis Brière (1962), Maximilien Rubel & Louis Évrard (1963), Émile Bottigelli (1971), Michèle Kiintz (1972), Corinne Lyotard (1973), Gérard Cornillet (1986). None of these deviates from the translation spectre, for Gespenst. The unpublished (ghost) translations of 1848, supposedly by Victor Tedesco and Hermann Ewerbeck, are probably apocryphal.

4 Engels makes this explicit in his letter to Sorge: “the French version especially is quite indispensable for the Latin countries of Europe as a counter to the nonsense purveyed by Bakunin, as well as the ubiquitous Proudhonist rubbish” (Marx and Engels, 1982: 343).

5 Ralph Waldo Emerson, “Self-reliance”, Essays, Cambridge, MA, Houghton & Mifflin, 1876, p. 52.

6 Marx-Engels’ term is one of the German words for a ghost. Unlike the more common cognate Geist, it bears a very close morphological and phonetic resemblance to the word Gespinst, which literally means “yarn”, but is used figuratively (just like the English) to refer to a “tall tale”, as in “to spin a yarn.”

7 Results of a comparative study of lexical collocations in political journalism for the year 1848 in The British Newspaper Archive and the BnF’s RetroNews.

8 These “image generators” are essentially designed to perform intersemiotic NMT.

9 While the general economic consensus has always suggested the threat of technological unemployment to be exaggerated, some AI experts disagree. Frey and Osborne (2017) suggest that almost half of contemporary US jobs are under threat from future computerisation.

10 The German original (Kafka, 1917 [1915]): Als Gregor Samsa eines Morgens aus unruhigen Träumen erwachte, fand er sich in seinem Bett zu einem ungeheueren Ungeziefer verwandelt. Er lag auf seinem panzerartig harten Rücken und sah, wenn er den Kopf ein wenig hob, seinen gewölbten, braunen, von bogenförmigen Versteifungen geteilten Bauch, auf dessen Höhe sich die Bettdecke, zum gänzlichen Niedergleiten bereit, kaum noch erhalten konnte. Seine vielen, im Vergleich zu seinem sonstigen Umfang kläglich dünnen Beine flimmerten ihm hilflos vor den Augen.

11 See Wang et al., 2022 for a synopsis.

12 See, for example, Baker, 1993; Laviosa, 2002; Kruger et al., 2013; Loock, 2016.

13 See, for example, Koehn, 2009; Callison-Burch et al., 2011; Koponen, 2012: House, 2014; Toral and Way, 2015; Doherty, 2017; Klubička et al., 2018; Läubli et al., 2018, Moorkens et al. 2018; Koponen et al., 2020; Popel et al., 2020; Popovic, 2020; Han et al., 2021.

14 Holz-Manttari’s theory of “Translatorial Action” (Schäffner, 2011) is broad enough to cover the possibility of a translator advising a prospective client not to translate a document at all (at least in the narrow sense).

15 There are obviously exceptions. Michael Cronin would be one notable example of a translation scientist who seeks to extend the discipline’s critical mechanisms into new global and political fields. See, for example: Michael Cronin, Eco-Translation. Translation and Ecology in the Age of the Anthropocene, London, Routledge, 2017.

16 See Steingard and Fitzgibbons, 1993. I am using TQM, inexpertly, as a loose umbrella term covering recent developments in quality control models such as Six Sigma, Lean Manufacturing, and ISO 9001.

17 http://prize.hutter1.net/hfaq.htm [consulted 10 February 2022].

18 https://www.youtube.com/watch?v=fn3KWM1kuAw [consulted 5 January 2022]

19 https://google.github.io/aichoreographer/ [consulted 20 March 2023]

20 This is a bravura piece of literary social-science writing, which playfully foregrounds the problematic intercultural transmission of Latour’s French theory by adopting an ironic American persona.

21 Michel Serres, Hermes III. La traduction, Paris, Les éditions de minuit, 1974. Latour and Serres worked closely together on the book Éclaircissements, Paris, Flammarion, 1994.

22 In this regard, I have been influenced by research-actors working in inclusive theatre translation, such as Elena Di Giovanni (2022), who provides an invaluable synopsis of the “inclusive turn”.

23 Unspoken Spoken, directed and choreographed by Fin Walker, Candoco Dance Company, 2018. https://candoco.co.uk/work/unspoken-spoken/ [consulted 8 February 2022]

24 https://teamyacht.com/projects/the-computer-accent [consulted 29 November 2022]

Haut de page

Pour citer cet article

Référence électronique

Samuel Trainor, « Mixing human and nonhuman dance partners: contrapuntal translation and the shift from replacement to inclusivity »Palimpsestes [En ligne], 38 | 2024, mis en ligne le 30 novembre 2024, consulté le 17 mars 2025. URL : http://journals.openedition.org/palimpsestes/9181 ; DOI : https://doi.org/10.4000/12spe

Haut de page

Auteur

Samuel Trainor

Sam Trainor est un traducteur et poète d’origine britannique. Il est maître de conférences en Traductologie à l’université de Lille. Dans sa recherche, il élabore une théorie du traduire polyphonique, en s’appuyant sur ses propres activités créatives. Sa dernière publication poétique est une retraduction de « La défloration de Lède » de Ronsard (PN Review 262, 2021), et il travaille actuellement avec des membres du LOOP (le Laboratoire ouvert sur l’œuvre poétique) sur la traduction en français de l’œuvre de plusieurs poètes britanniques contemporains. Il a publié de nombreux articles et chapitres d’ouvrage sur divers aspects de la traduction littéraire, qui développent tous les notions de la polyphonie, la syncope (musicale) et la multimodalité. Il a également co-dirigé un ouvrage collectif sur la traduction culinaire, à paraître d’ici la fin 2025. Il est l’un des rares chercheurs à avoir publié une étude sur la traduction professionnelle du scénario filmique : « Cinema skopos: Strategic Layering and Kaleidoscopic Functionality in Screenplay Translation » (Palimpsestes 30, 2017). Hormis ses activités littéraires et traductologiques, il a traduit plusieurs articles et ouvrages scientifiques, dont Darker Shades: the Racial Other in Early Modern Art de Viktor Stoichita, paru chez Reaktion Books en 2019.

Articles du même auteur

Haut de page

Droits d’auteur

Le texte et les autres éléments (illustrations, fichiers annexes importés), sont « Tous droits réservés », sauf mention contraire.

Haut de page
Rechercher dans OpenEdition Search

Vous allez être redirigé vers OpenEdition Search