Corps de l’article

Writing in 2005 on the emerging conceptual relationship between translation and society, Daniel Simeoni addresses what he perceives as a striking enigma in the social sciences. Why have the social sciences had so little to say about translation?

The question of “translation and society” has not always been a topic on which research could be done; it has not always been possible to write on the links between the two concepts, on the role played by translations and translating in society, on the social dimensions of the practice, the interplay of the complex forces shaping the politics of translation worldwide, or on the history of these interrelations

Simeoni, 2005, p. 3

The silence for Simeoni is all the more puzzling given the relative inclusiveness of the sociological eye. Why was translation not included among the legitimate objects of sociological enquiry at an earlier stage when even the most private acts (e.g.: suicide) were deemed fit for investigation and almost every branch of the human sciences, including notably history, underwent change as a result of the work of Durkheim and his followers (ibid., pp. 8-9)? His conclusion is that fundamentally social scientists were not sufficiently self-reflexive to free themselves from their national contexts; “the cognitive confinement within which scholars operated was due to the fact that the words and the rhetoric they used, the ways in which the new treatises and analyses were elaborated, followed what were typically national (nationalistic) traditions of thought” (ibid., p. 9). In this context, the true scandal of translation is translation itself, which, although it underpinned the working practices of the polylingual pioneers of the social sciences, could not be foregrounded as it would bring to light the nationalist or situated origins of social theory and analysis. For Simeoni, this state of affairs was not a regrettable incident from the past but a persistent feature of the present and translation had still a long way to go before it would become a core concern, for example, of North American Anglophone social sciences. In this essay, I would like to continue to bear Simeoni’s question in mind but parse his answer somewhat differently. In other words, I would like to examine how the universalism which underpinned the claims of the early social theorists about the forms of social life, re-emerges in contemporary paradigms of automated translation. I would further like to argue that Simeoni’s contention that translation foregrounds the situated, contingent basis of universalist claims is borne out by developments not only in the past history of translation but in the current industry of translation.

Friction

Digitalization was not a core concern of Daniel Simeoni’s research but it is striking how the tensions in the conceptual relationship between translation and society he analysed are working their way through the contemporary moment. In an article that first appeared as a Xerox PARC Working Paper in 1980, Martin Kay claimed that “history provides no better example of the improper use of computers than machine translation” (Kay, 1997, p. 6) and he went to imagine a scene of sorry hubris:

There was a long period—for all I know, it is not yet over—in which the following comedy was acted out nightly in the bowels of an American government office with the aim of rendering foreign texts into English. Passages of innocent prose on which it was desired to effect this delicate and complex operation were subjected to a process of vivisection at the hands of an uncomprehending electronic monster that transformed them into stammering streams of verbal wreckage. These were then placed into only slightly more gentle hands for repair.

ibid., p. 7

Kay believed in the possibilities of machine translation but he did not believe in its impossible ambitions. He described translation as a “fine and exacting art” but “there is much about it that is mechanical and routine.” If these mechanical and routine parts could be given over to machines, translators would not only be more productive but the work would become “more rewarding, more exciting, more human” (ibid., p. 3). The problem, of course, was deciding what it was that was mechanical and routine. Kay’s main critique of approaches to MT was that they favoured solutions that were approximate and tended in the absence of human intervention to multiply errors in catastrophic chain reactions. So it was assumed that if a machine translated a pronominal reference correctly 90% of the time, this was an acceptable outcome. The problem is: how does the machine know that it is translating the pronominal reference correctly or not? If there is no reliable way of knowing which 10% have been incorrectly translated, then 100% of the pronouns must be examined by the human editor or translator. As Kay observed, “it does not matter very much if the program is right 90, 99, 80, or 50 per cent of the time. The amount of work that it leaves for the repairman is essentially the same” (ibid., p.10). Kay’s proposal, the “Translation Amanuensis”, a Translation Editor that worked in conjunction with rather than trying to replace the human translator, became the template for the forms of computer-assisted technology that have come to dominate the translation profession in the years since Kay’s working paper was first published.

What is significant in Kay’s critique is the nature of his criticism. He sees the fallacy of particular approaches to machine translation lying not in the entirely reasonable ambition to automate certain sub-routines in the translation process but their indifference to details. The question of details, the translation of a pronoun, for example, is related to reliability: “If it [translation system] falls short of the acceptable standard, to any degree whatsoever, it might as well fail grossly because the burden it places on the proofreader will be very large” (ibid., p. 11; his emphasis). He sees the role of the human translator interacting with the Translation Amanuensis as having primarily to do with a close attention to linguistic and translation detail so as to prevent the “cascading errors” that are all too common in “language processing” (ibid., p. 22).

This notion of “detail” takes on a wider significance if we situate it within Simeoni’s concern in his 2005 essay about the failure of the social sciences to attend to the “details” of language and translation: “[t]he idea of an interpenetration of cultures, high or low, including their own understanding of foreign practices which they viewed as evidently social, completely evaded writers and readers alike” (Simeoni, 2005, p. 10; his emphasis). Jean-Claude Milner in a discussion of Walter Benjamin sets up a distinction between “thinking in a massive way” (penser de manière massive) and “thinking in details” (penser en détails) (Milner, 2011, p. 31). For Milner, one of the pitfalls, for example, of progressive thinking is the “rhetoric of massiveness”, the tendency to employ specific terms with a supposedly mass or broad effect such as “freedom”, “democracy” or “empowerment” in ways that silence the hearers. If details such as mass fingerprinting at ports of entry or the decreased importance of parliamentary accountability through the rise of executive power seem to contradict the effective purchase of these terms, they are readily dismissed as mere details with respect to what is seen as the defence of the more fundamental achievement of parliamentary democracy itself.

On a darker but analogous note, a frequent claim of French negationists has been to dismiss the Holocaust as a “detail”. As Milner claims, “[w]hen a subject notices a detail, no matter how small, and is told to ignore it, he or she can be sure that something very important is going on there[1]” (ibid., p. 32). What thinking in a massive way results from is described by Milner as the “universel facile” whereas “if thinking through details leads to the universal, it is of necessity a difficult universal.[2]” (ibid.) For Milner, Freud’s psychoanalytic explorations in the Traumdeutung or the Psychopathology of Everyday Life are examples of detailed thinking which take massive notions like hysteria, neurosis and psychosis and track them through a series of illuminating details. At the end, the notions may still bear the same name but their internal structure and the nature of the universal claims they make have radically changed. The difficult universal is arrived at not by systematically identifying what each case has in common as each case is very different but by moving towards a notion of emergent commonality based on difference rather than similarity. In order for this to happen, of course, Simeoni’s “interpenetration of cultures” has to be brought to the fore to both delineate the lines of difference and understand how translation contributes to any notions of emergent commonality.

Alan Melby in his “Notes” on Martin Kay’s classic paper sets up a distinction between two text types that produce very different kinds of results in machine translation. The first text-type is “controlled domain-specific language” and the second is “dynamic general language” (Melby, 1997, p. 29). Controlled domain-specific languages are languages where lexicon, syntax and forms of expression and reference are tightly controlled. In these cases, controlled domain-specific languages MT systems are capable of producing high-quality raw output that require relatively little post-editing. This is not the case, he argues, with dynamic general language texts. The results are frequently uneven and disappointing and require an advanced degree of human post-editing to bring them up to the standard of high-quality output. This leads him to a second set of distinctions where he sets up an opposition between high quality and what he terms “indicative translation”:

It has often been assumed that for a translation to be useful it must be of sufficiently high quality to be comparable to the work of a professional human translator. Not so. Low-quality MT that is produced quickly and used only to get an indication of the content of the original text and which is then often discarded is sometimes called “indicative translation.” Surprisingly, indicative translation is perhaps the fastest growing use for MT.

ibid., p. 29

In the case of indicative translation, what one is primarily concerned about is the overall or “massive” effect of the text. As Melby notes, “who cares about grammar or word choice when a motivated human can, with a little practice, form an approximate idea of what the document is about?” (ibid., p. 30). The details are unimportant, it is the overall effect that counts. The proliferation of translation applications for smartphones in addition to the now almost axiomatic invitation to translate when Google searches throw up content in foreign languages are powerful multipliers for the practice of indicative translation. Implicit in the widespread availability of automatic translation is a notion of translation as potentially instantaneous and universal. This availability is of course facilitated by changing approaches to MT and the relentless increase in the processing capacity of computers themselves.

Until the late 1980s MT was largely dominated by rule-based systems where grammar and syntax rules were combined with cross-language dictionaries. In the 1990s the shift was to experimenting with sets of parallel texts. In statistical-based MT, algorithms analyze large collections of previous translations or parallel corpora to estimate what the statistical probabilities are of words or phrases in one language ending up in another. A model is then constructed on the basis of these probabilities and used to evaluate new text. By implication, these systems perform best on the types of texts on which they have been trained. The greater the coverage, the greater the need for more and more extensive corpora (Ratliff, 2006, pp. 2-3). The paradigm implicit in the statistical approach is one of analysis of massive amounts of data available on the Web or elsewhere. One could argue, in effect, that it is an example of the particular kind of effectiveness of “thinking in a massive way.” The result is that in both of Melby’s oppositions we can see the emergence of the easy universal that Milner describes. In the case of controlled domain-specific language we have what might be described as a pre-emptive universality. One of the goals of controlled language is to remove “accidental content”, the many different ways in which the same thing can be said and which generates new content to be translated. This “accidental content” is the details that must be removed if the MT system is to run effectively. Once the language is sufficiently controlled then the universal rolling out of the translation in the language pairs catered for is eminently feasible.

In the case of indicative translation, the details are not removed, but they are no longer deemed to matter. It is the cumulative effect of meaning based on human ingenuity in dealing with less than intelligible texts that legitimates the roll-out of online translation services on smartphones or laptops. The implicit goal is the vision of the universal translator described by Evan Ratliff in his discussion of a new MT approach developed by Jaime Carbonnell, science officer with the IT company, Meaningful Machines:

Right now, the Global Autonomous Language Exploitation program run by Darpa [Defense Advanced Research Projects Agency] is aiming to complete an automated text and translation system in the next five years. Meaningful Machines is part of a team participating in that challenge, including the “surprise language” segment (in which teams are given a more obscure language and asked to build a translation system). The challenge sounds like another attempt to create the sort of universal translator that has eluded MT for 60 years. But success seems much more plausible than ever before.

Ratliff, 2006, p. 2

It is important to understand that the notion of easy universal has nothing to do with the very considerable technical complexity involved in the construction of controlled natural languages or the development of MT systems underlying the provision of online indicative translation. Rather, the difference lies in the role of detail and the relationship between translation and detail in positing another notion of universality.

Although he does not explicitly address the question of the digital, Simeoni contests the notion of easy universalism and the telling disregard for the detailed transactions of translation:

Proper translation, as has been amply demonstrated in the restricted field of translation studies over the last twenty years or so, is never simply a replica. An appropriate dose of “friction,” in the sense of being neither too aggressive nor too ignorant of the other, is inevitable, giving rise to mutual misunderstandings as an ingenious solution to ordinary, yet potentially devastating disagreements in social life.

Simeoni, 2005, pp. 13-14

It is the presence of this “friction” deemed by Simeoni to be inescapable, “inevitable,” the daily working out of the detailed difference of language and culture in translation, which can give rise to potentially devastating consequences in social life. It is arguably the awareness of the devastating consequences of detail which explains why there is such a thing as a translation profession in the first place and why students will spend years acquiring the requisite language and other skills to become translators. It is the attention to detail that is seen time and time again to characterize the competent translator. When translators give voice to a characteristic if not always enabling modesty about what they do, it is often in the awareness of the sheer enormity of detail that crowd into the rendition of a text.

Eleanor Marx in her introduction to the 1886 and 1892 edition of her translation of Madame Bovary claims: “Certainly no critic can be more painfully aware than I am of the weaknesses, shortcomings, the failures of my work” (Flaubert, 1886, p. xxi). As a translator who becomes her own critic, she knows that the painful awareness lies in the multiplicity of detailed decisions or choices she has had to make to bring Flaubert’s text to the English reader. She also anticipates in a way the nature of translation criticism which focuses on the aesthetic, cultural and political implications of the choices that are made at micro-levels by the translator. When Vladimir Nabokov subjects Marx’s translation to the withering ire of his analysis, he singles out one tense, the imperfect tense, crucial to conveying a sense of unity and continuity in time, as a significant detail that weakens the force of the translation (Nabokov, 1980, p. 173). Emily Apter in her discussion of Marx’s translation sees a political philosophy underpinning certain lexical choices by the daughter of Karl Marx and a significant activist in her own right. Whereas Alan Russell in his later 1952 Penguin translation consistently translates Flaubert’s “la richesse” by “riches”, Eleanor Marx always renders the word by “wealth.” Apter argues, “Eleanor Marx’s consistent rendering of ‘riches’ as ‘wealth’ would seem to enhance the latent critique of wealth in Flaubert” (Apter, 2008, p. 75).

To paraphrase Milner, if the subject notes a translation detail, no matter how small, and is told to ignore it, he or she can be sure that is where something is going on. There is no escaping, as Simeoni would phrase it, the “friction” of translation. It is indeed the necessary care for and attention to detail that makes research in translation studies arduous and time-consuming (see, for example, in the case of the multiple English translations of a sole Jules Verne text O’Driscoll, 2011).

At another level, what is implicit in the example of translation production and analysis discussed above is the presence of the difficult universal, the universal that is arrived at through the enumeration of difference rather than the re-statement of commonality and which problematizes the “massive” notions of fidelity, equivalence and meaning. There is a sense then in which the easy universalism of particular representations of translation in the digital age runs directly counter to another form and reception strategy of translation based on the idea of a difficult, asymptotic universal. This difficult universal recognizes that there is a communicative life beyond particulars or details, otherwise the translation enterprise would be doomed to a plaintive solipsism, but that any move towards the universal must take account of the endless interrogation of the details of language and culture. Lawrence Venuti argues that the “political intervention performed by translation in postmodern culture may be more usefully imagined as a local, small-scale activity of resistance against dominant discourses and institutions” (Venuti, 2008, p. 22).

In his discussion of the French translation by C.A. Alexandre of the fifth volume of Theodore Mommsen’s Römische Geschichte, Daniel Simeoni notes the hostile preface of the French translator to the subject matter of his translation and claims that in this, “Alexandre was following a long tradition of ideological control or State-induced reconstruction by translators of histories by foreign authors” (Simeoni, 2005, p. 10). If in the aftermath of the Franco-Prussian conflict, translation or rather translation commentary became war by other means, this confirms for Simeoni the nationalist recasting of translated exchanges in European social sciences. It is useful to situate Simeoni’s observation in the context of the nature of institutional or political investment in the universalism of MT. As we saw above in the reference to the Global Autonomous Language Exploitation program, the Defense Advanced Research Projects Agency, the research wing of the US military, is particularly interested in the possibilities of machine translation. Language Weaver, a Californian firm that specializes in statistical MT systems, “got an investment from the CIA’s venture firm In-Q-Tel in 2003” and “now has customers in intelligence agencies here and abroad” (Ratliff, 2006, p. 3). In the words of the CEO of Language Weaver, Bryce Benjamin, the software “is being used day in and day out to catch bad guys” (ibid.). The ultimate political problem, of course, is in defining who the “bad guys” are and for whom. More significant, perhaps, than the cavalier moral dualism of the comments, is that from the Cold War to the “War on Terror”, a certain kind of universalist vision of translation has accompanied national military bids for strategic and territorial advantage. It is as if the massive thinking behind the tactics of Shock and Awe were seeking a congenial home in the field of translation and that a theory of global power and influence cannot forgo a theory of what translation should be and do. If the absence of translational self-reflexivity in social sciences pointed in Simeoni’s views to the enduring presence of national conflict, it is striking that the persistence of conflict fuels not so much a disavowal of translation as a reframing of the activity within a default universalism of influence and control.

Borderlines

Re-instating the agent as a socialized subject in translation and in translation studies was a recurrent preoccupation of Daniel Simeoni. He was particularly alert to the difficult positioning of the translating agent as nomad, “[t]he translating agent straddles the borderline between cultures. Although various pressures associated with practice force him/her to ‘stay home’—on the target side—s/he cannot afford to ignore the source-field a long time without being at risk” (Simeoni, 1995, p. 453). In developing an approach to translation predicated on the difficult universal, the question that might be asked is, what for? In a digital world of accelerated, quasi-instantaneous exchange is there not a compelling case to be made in terms of time and efficiency for the easy universalism of controlled language and indicative translation? Is there not a sense in which some communication, however imperfect, is better than none? Is not the view from the digitalized translation agent, the agent instituted by the automated translation paradigm, that more is better, and more, more quickly is better still? To try and answer this question, it is helpful to ask what it is we put into communication when we translate. In the case of interlingual translation, we might respond that it is two languages and cultures that are typically placed in dialogue, and there may be sublanguages of these languages (legal, scientific, commercial) or subcultures (based on, for example, ethnicity or class) that are party to this dialogue. Looking more closely at the word dialogue, there are two components, dia from the ancient Greek dia meaning “across” and logue related to legein, “speak”. Implicit in the term is the acknowledgement of a distance that needs to be crossed (dia) and the possibility that the distance can be crossed through the intelligibility of speech (legein). Therefore, translation as a dialogue between languages and cultures must be as much about the necessary recognition of distance as it is about the communicative possibilities of intelligible speech, oral or written. This is expressed slightly differently by Simeoni when he has recourse to Norbert Elias’s distinction between involvement and detachment (ibid., p. 448). Elias contends that the physical sciences have been well served by a detached approach to describing or analyzing physical events but that in the human sciences, active participation and involvement in the feelings and experiences of human subjects is an absolute requirement if anything of significance is said about human actions (Elias, 1956, pp. 226-252). What Simeoni sees as desirable for a proper structuring of the discipline of translation studies, a blending of detachment and involvement that is hinted at by the notion of phenomenological structuralism in linguistics (ibid., pp. 449-453), is implicit in the dialogical understanding of the act of translation. How we might give effect to this blended method demands that we re-examine the key concepts of difference and identity in culture.

The notion of difference is stock-in-trade of any debate on intercultural communication and is a powerful vector for one of the world’s major economic activities, tourism. However, a difficulty arises when the so-called typical differences (Japanese formality, American informality, Caribbean joie de vivre) become a barrier rather than an aid to understanding as they harden into the exportable cliché of mass tourism or the semiotic shorthand of commercial soap operas. As the sinologist and philosopher François Jullien argues:

Between cultures, I would not trust these supposedly characteristic differences, labeled as such and presented as standard (the most obvious traits are often the least interesting): as they become ossified, they become an obstacle to thought. But, I said it before, I make the gaps work—the notion is exploratory not classificatory. Opening up a gap is to break with conformism, to bring tension back into thinking, in short, to set our reason back to work.[3]

Jullien, 2010, p. 15; his emphasis

The notion of “gaps” here is as much within as between cultures. In other words, cultures are not uniform blocs reified under the sign of difference which are assimilated by translators and then bridged by their irreproachable sense of tact. They are dynamic entities, constantly in a state of flux. For this reason, the notion of “identity” becomes highly problematic as the question is what kind of identity are we talking about, given that any given culture or language is a product of endless mixing and cross-fertilization and that new ways of working, generational change, new forms of technology, subject the language and culture to continuous transformation. In this context, Jullien argues not for the promotion of “identity” which, in many cases, is, to a greater or lesser extent, fictive but for the idea of “fecundity” (fécondité) (ibid., p. 12). Fecundity carries within it a dynamic sense of plurality that foregrounds the resources (ressources) of a culture. The notion of “resources” here is not to be confused with that of “values,” [“values are the vectors of an affirmation of self. They are bound up, whatever one might claim to the contrary, in a relationship of power whereas resources are indefinitely exportable (exploitable) and available to everyone”[4]] (ibid., p. 15). Confucianism, for example, offers the thinker the resources of subtlety of expression, sense of balance, the importance of a notion of “regulation,” the avoidance of overly dogmatic thinking but as a value system, Jullien argues, it can be less attractive in promoting social conformism, a servile attitude towards those in power and so on (ibid., p. 16).

The form of difficult universalism that works through the details of language and culture in translation both reveals the “gaps”, that distance that needs to be crossed in translation, and the resources which are made available to world languages and cultures through the translational circuits of intelligibility. In order to see how we might relate this form of universalism to translation practice, we will examine two very different accounts of what it is to translate.

Meghaduta or “Cloud Messenger” was a narrative poem in Sanskrit written by the famed poet Kalidasa in the fourth or fifth century. It was translated into English in 1813 by Horace Hayman Wilson. Although sympathetic to the aims of Empire, Wilson was at pains to point out how different was the world represented in the Sanskrit poem. As David Damrosch notes, “he follows the sixty pages of the poem with over a hundred pages of detailed and informative notes, explaining religious and geographical references, the symbolic significance of birds and plants, and the social and literary assumptions of the poet and his audience” (Damrosch, 2008, p. 44) What Wilson is paying attention to in his copious notes are the “details” of Sanskrit language and culture that inform the poetic production of Kalidasa, the necessary “gaps” in Western attempts to grasp the poem. He pursues the project of difficult universalism using a dual strategy of repudiation and analogy. The repudiation is the direct challenge to Western stereotypes about Indian culture, stereotypes that he feels are undone by the import of the poem. When Kalidasa refers to the virtue of gratitude, Wilson notes:

The Hindus have been the object of much idle panegyric, and equally idle detraction; some writers have invested them with every amiable attribute, and they have been deprived by others of the common virtues of humanity. Amongst the excellencies denied to them, gratitude has always been particularized; and there are many of the European residents in India, who scarcely imagine that the natives of the country ever heard of such a sentiment. To them, and to all detractors on this head, the above verse is a satisfactory reply.

Wilson, 1814, p. 91

In his comment, Wilson is undercutting a notion of typical difference and opening up a gap in received thinking about Indian culture. The other component of Wilson’s strategy is analogy where he refers to classical writers of antiquity such as Ovid, Catullus and Horace to make Kalidasa comprehensible to his Western readers. After the “dia” of distinctive difference, he moves towards the “logos” of intelligibility, the attempt to make the meanings of the poem circulate in his culture of origin:

the analogies between the poetry of the East and the West, are given especially for the benefit of those liberal critics, who admire, upon the strength of prescription, the beauties of classical and modern writings, and deny all merit to the same or similar ideas, when they occur in the works of oriental writers. It is also entertaining to observe, how much men resemble each other, in spite of the accidental varieties of complexion or education, place or time.

ibid., pp. xix-xx

Wilson’s move here towards a notion of the universal is not based on the repudiation of difference or an attempt to fill in the gaps but on a repeated desire to open up a breach in conformism, bring tension back into Western perceptions and challenge the foundations of Empire’s way of “reasoning” about its colonial subjects.

This is not to argue that Wilson was not in many other ways wholly complicit in the project of Empire but rather to show how his translation enterprise is founded on a notion of “thinking through details” that eschews by and large the “massive” effects of colonial stereotyping. Unfortunately, for Wilson, his “logos” was not quite up to the task of translating the Meghaduta into English. Captivated by the translation poetics of neo-classical writers like Dryden and Pope, Wilson renders the poem into English in less than heroic couplets. As Damrosch notes, “[t]ranslations notoriously age as their language becomes dated, but Wilson’s style was dated even in his own time” (Damrosch, 2008, p. 45). What Wilson reveals in his detailed presentation of the poem is not so much a static notion of Sanskrit identity as an image of the abundance of “resources” in Sanskrit culture and language. Indeed, it is precisely in terms of a resource-oriented approach to intercultural contact that Wilson presents his translation project in the preface:

The efforts of Sanscrit scholars have hitherto, however, been directed rather to the useful than the pleasing, rather to works of science than imagination. The complicated grammar of the Hindus has been most successfully investigated, their mythology amply illustrated, and much of their philosophy satisfactorily explained; their astronomical works have been exhibited to the philosophers, whose modern attainments have rendered ancient science an object of curiosity rather than information, and their laws are no longer concealed behind the veil of an unknown tongue, from the knowledge of those who are charged with the administration of justice in Hindoostan. It only remains, to explore the field of their lighter literature, and transfer some of its most elegant flowers to a European soil.

Wilson, 1814, pp. x-xi

Wilson’s comments, of course, are replete with a particular kind of imperial hubris (“successfully investigated”, “amply illustrated”) and he is not averse to spelling out the coercive possibilities of knowledge (“administration of justice in Hindoostan”). On the other hand, as revealed by his prefatory comments and his detailed annotations to the translation, he is interested in the multiplicity rather than the unity of the culture he investigates. What is made apparent, above all, is the fecundity of the culture and language which informs the Meghaduta. The question that might be asked however is what relevance does this particular form of “thick translation” by a 19th century British Orientalist have for translation in the 21st century?

What is paramount in Wilson’s approach to his translation of the Sanskrit text is an attention to and a concern with detail. This indeed is often seen as an integral part of how translators go about their business. Daniel Gouadec claims that “[t]ranslators must first and foremost strive to avoid making serious errors” and he gives as examples, “mistranslating drug dosages, switching around the connections in a wiring diagram, confusing a rise with a fall or clockwise with anti-clockwise” (Gouadec, 2007, p. 10). Of course, what might seem like a detail, a missing zero or a misplaced term, could in all of these cases have dramatic consequences. However, the concern with detail in the digital age should not simply be reduced to the War on Error. This is because the nature of what might be considered “detail” is changing and this is related to the reconfiguration of two basic parameters, time and space.

Dominique Estival, in a discussion of the development of a language translation interface for the Australian Defence Organisation (ADO), describes how a greater spatial or geopolitical sensitivity and an increasing concern with timescale, meant that translation became a major preoccupation for the ADO. Crucially, it is the nature of contemporary conflict that motivates the move towards translation; “the shift of focus from ‘defence of Australia’ to ‘national security’ implies an increased awareness of the international environment around Australia” (Estival, 2005, p. 178). The three activities that result from the shift in the defence paradigm are intelligence gathering, coalition operations and foreign operations (peace-keeping, humanitarian and relief operations). For example, in relation to intelligence gathering, the implication of translation in a New Intelligence Order is made particularly clear:

There have been many discussions for better and more timely intelligence since the intelligence failures shown to precede the tragedy of 9/11 in the USA and requests for more translators and tools to help translators have been widely publicised. Australia is in the same situation as all other countries in this respect, although the Bali bombings in September 2002 and September 2005 and the bombing of the Australian embassy in Jakarta in October 2004 mean that there are also specific threats and concerns for Australia with particular linguistic implications.

ibid.

The notion of intelligence vulnerability which is universalized (“Australia is in the same situation as all other countries in this respect”) means that the notion of intelligence itself, rather like information, cannot be thought of outside the operation of translation. What “increased awareness of the international environment around Australia” leads to is a greater engagement with the linguistic detail of the Asia-Pacific region. In other words, whereas previously traditional ties to the UK and US had led to English being the sole language of use of the ADO, different political circumstances and new allies (Japan, Republic of Korea, Thailand, France) as well as new areas of non-combat operations (Solomon Islands, East Timor, Aceh (Indonesia)) meant that not only was there a greater contextual sensitivity to language use but there was also an acknowledgement of growing complexity. The default universalism of English as a global lingua franca no longer functioned. Translation was as inevitable as it was necessary. If translation was necessary because of changing geo-political or spatial orientations, what were the implications for time?

According to Estival, it takes about one to two years to train someone to function in a spoken foreign language and another two to three years to produce an effective translator/interpreter depending on the language pair. He adds, “it is very difficult to predict which languages are going to be of interest in a three-year time frame and even more difficult to predict the extent of the potential demand for translation for those languages” (ibid., p. 180). Given the size of the Australian population, the ADO cannot be increased beyond a particular size so that there “will never be enough personnel available to be trained and the range of languages of interest cannot be predicted in time to perform the training required to produce skilled translators in those languages” (ibid.). Estival is drawing our attention to is what we have commented on before, namely the question of chrono-diversity, the tension between the short timeframe of translation demand and the long timeframe of translator training or education. Apparent in the Australian example as elsewhere is that a shift in scale at a spatial level has immediate consequences at a temporal level. As the linguistic demands, in a sense, become more detailed and more complex, the ability to deliver on them in real time becomes more and more problematic.

The response of the research wing of the Australian military, the Australian Defence Science and Technology Organisation (DSTO), to this scalar shift was to develop a Language Translation Interface (LTI) that brought together the resources of existing MT systems, including free, online MT translation services. Time and money are invoked as the principal reasons for the development of the LTI; “The development of a new translation engine requires enormous efforts and resources and is beyond the scope of a research project at DSTO. In any case, it is not possible to predict which languages might become of interest and the results of such efforts would most likely not meet actual needs” (ibid., p. 189). At one level, the results and the basic philosophy of the LTI would appear to fall under the rubric of the easy universalism that we have seen informing the MT translation projects of the US military. The need to think in detail about new geo-political situations leads to a form of “massive” translation through the widespread use of available MT tools. The overall conclusion in the Australian case was that the users were broadly happy with the kind of indicative translation output provided by the LTI.

The translation devil does, however, intervene at the level of detail. In the one example provided of output from a LTI session, where a number of alternative English translations are provided for a Japanese sentence, it is clearly the MT output that has been post-edited which provides the most satisfactory translation. The repeated references to “quality” and the desirability of developing and building Translation Memories to enhance the quality of the LTI output show how the concern with detail is not so much removed as displaced. In an age of the widespread deployment of IT, there may increasingly be a sense in which translation or what we tend to think of as translation may not be where we expect find it and which indeed it may, in fact, a lot of time be going under another name. The concern with detail which is part of that move towards difficult universalism may not be an inevitable casualty in the digital age and may re-emerge in a different guise.

Discriminating Tasks

In his analysis of a lecture given by the British linguist John Lyons in Tours in 1983, Daniel Simeoni draws attention to the nature of the phenomenological structuralism described by Lyons. The primary task of a linguistics underwritten by phenomenological structuralism would be the modal component of the clause, in other words, “everything that signals the author/translator’s involvement in the formulation (énoncé) as opposed to its propositional content, or the part logically processed” (Simeoni, 1995, p. 452). For Simeoni, it was the modal component that would continue to act as the nemesis for “translation machines,” the test they would always “flunk.” He goes on to remark:

Machine translation has been, and will remain notoriously deficient in dealing with the modal component of énonciation, for principled reasons. A large part of the difficulty of translating can thus be theorized as a “discriminating task”: sifting the idiosyncratic character of the utterance out of the instituted componentiality of the translating agent’s production.

ibid., pp. 452-453

In the intervening period since Simeoni formulated his observation machine translation has, of course, evolved. However, what is more significant in the light of Simeoni’s observations is how his remarks on the “discriminating task” of translation have not so much been ignored as folded into issues around post-editing and concerns with respect to translation quality. In the 1980s and 1990s the use of post-editing on machine-generated text was largely for the purpose of making the translated text minimally comprehensible. The text was intended for dissemination rather than assimilation akin to the indicative translation mentioned above by Alan Melby. As Garcia notes, “[f]ull post-editing was considered to involve more effort than translating directly from the source text” (Garcia, 2011, p. 218). However, as MT systems improved and became more widespread in their use, “full post-editing is now encroaching into areas that had been dealt with up to now by translation assisted by TM [Translation Memory]” (ibid., p. 218). Two factors are behind this development, both relating again to space and time. The global connectivity of economic activities means that not only is economic power shifting with the emergence, for example of the Chinese, Brazilian, Indian and South African economies, but the demand for certain language pairs such as Chinese-English has risen dramatically. Given the paucity of English mother tongue speakers with an adequate command of Chinese, the tendency is for Chinese language speakers to translate from English into their mother tongue.

Garcia argues in a study of the effectiveness of post-editing that it was translators working into their weaker tongue that tended in particular to benefit from post-editing MT output as opposed to translating directly from the source language: “should the quality be high enough, the MT version may save translators time in the process of understanding the source and provide them with a draft on which to work” (ibid., p. 221). Spatial reconfiguration of economic relations means increasing traffic between non-cognate languages with consequent translation challenges. At the same moment, as economies in the information age are connected in real time, the response times to demands of various kinds are endlessly foreshortened. In this context, TAUS, a think tank for the translation industry, claims that post-editing for publication should be able to process about 5,000 words per day, which is twice as much as conventional translation, even if they admit the quality is likely to be lower. The other caveats are that the figure supposes experienced, professional post-editors working on the output of engines trained in domain-specific areas and with the possibility that some of the text has been pre-edited (TAUS, 2010, pp. 6-9).

What is striking in the discussions around postediting and translation automation generally is the recurrent concern with “quality.” In a report on a TAUS Executive Forum held in Japan in April 2012 Jaap van der Meer noted: “The adoption of MT technology makes translation more efficient but what about the quality? This question creeps [sic] up in every presentation of machine translation solutions of course. We dedicated a special session to the problematic area of translation quality evaluation” (van der Meer, 2012, p. 2). Quality is, in a sense, the return of the repressed translation detail. The careful, detailed attention to text, language and meaning that is implicit in the act of translation re-emerges in the context of automation in the debates about the extent and role of post-editing and how to achieve acceptable quality in translation output.

The “massiveness” of the move towards automation is repeatedly wrong-footed by the detailed susceptibilities of “quality,” the challenges of Simeoni’s “discriminating task.” Haunting discussions around translation automation in the digital age is the spectral presence of the thickness of detail that make up languages and cultures. This is not to say that translation automation is not possible or highly effective in certain circumstances. It is and its use and practice will continue to grow for the spatio-temporal reasons we described above. However, it is important to note that translation is a scalar concept which covers a wide variety of practices and that our expectations of what it should or can do can ultimately be traced back to competing versions or understandings of the universal.

In an essay published posthumously on the “geopolitics of translation theory,” Daniel Simeoni returned to his interest in the contingent origins of theory, situating the work of polysystem theorists and Bourdieusian sociology in a specifically European notion of the state and state development. He uses the term “cultural loyalty” to characterise “the researcher’s internalized preferences for homogenous groupings representative of the culture under study, more often than not his or her own” (Simeoni, 2008, p. 337). He goes on to claim that:

This loyalty has often taken the guise of a theoretical agenda modeled after the cohesive strength entailed in European state building (but where is the alternative model, even today?). Few scholars are aware of this connection, so much so that much of the work that goes on in the discipline of the social sciences and the humanities follow traditions closely linked to the development of their institutions, that even as they question them, may in fact be steeped in a geopolitical unconscious, l’impensé géopolitique de la théorie.

ibid.

In Simeoni’s view, theories that claim to transcend borders are often inescapably defined by them. In particular, there are forms of the universal that seek to repress this geopolitical unconscious. What these are and how they might illuminate Simeoni’s claim can be illustrated by attending to a distinction made by a theorist cited in Simeoni’s essay, Jean-Claude Milner.

In a discussion of Jean-Jacques Rousseau’s famous declaration at the beginning of his Social Contract, “L’homme est né libre et partout il est dans les fers” [Man is born free and everywhere he is in chains], Milner claims that the opposition appears initially to be chronological. Man is born free at birth and then he becomes enslaved. Milner agues that a deeper opposition resides at a logical level:

I hear the sound of the clash between the universal proposition in the singular, Every man is free and the proposition in the plural, All men are free. The first one is true, the second false. But, at the same time, we understand that the proposition in the singular is only true in an intensive sense. It is universal in the strict sense that it brings out the maximum intensity in the name man. It would still be universal even if men were nowhere to be found free.[5]

Milner, 2011, p. 36

The kind of “intensive” universality evoked by Milner where there is an exploration of the maximal meaning or meanings of what a word might signify contrasts with an extensive universality primarily concerned with extension and plurality, as in mass consumer products, where what is most characteristic is their interchangeability and omnipresence (the Starbucks phenomenon). Translation in the digital age is faced with the tension between forms of extensive universality that drive the translation industry worldwide and the claims of intensive universality which underlines the maximally difficult and maximally complex nature of words and their use. The tension is referred to by Brian McConnell (quoted in Joscelyne, 2011, p. 1) from the software company Worldwide Lexicon Inc., who recommends a watchword for translation into future, “Don’t let perfect be the enemy of the good”:

His example of a model platform for a world of ubiquitous translation functionality where simplicity is the watchword is Twitter. The focus should be exclusively on defining conventions for the most common tasks and interactions between the various [sic] involved, and then regularly improving them.

Joscelyne, 2011, p. 1

If the “perfect” is the drive towards intensive universality, the “good” as defined here is the move towards extensive universality where the prior definition of conventions will allow for the cheap, fast and efficient circulation of messages in a “world of ubiquitous translation.” In a world of ubiquitous translation, however, it may be that the only sustainable good is the imperfection of the detail that alone can trouble the complacent universalism of the powerful. In this context, it is worth situating Simeoni, to use his own term, as a particular kind of “scholarly agent” (Simeoni, 1995, p. 452) in translation studies. Citing a work of Milner that deals with the condition of European Jewish intellectuals (Le Juif de savoir (2006)), Simeoni notes that “they never fully integrated into their host countries” and that their “experience was not diasporic; it was exile” (Simeoni, 2008, p. 333). Simeoni’s description of the situation of those European Jewish intellectuals who went into exile in North America could be read mutatis mutandis as a nod to his own biography. There is a distinct sense in his writings on translation that his concern with the view from the agent, the presence of the geopolitical unconscious, the attention to the discriminating task of the translator and the preoccupation with the social dimension to the activity of the translator is animated by that recurrent sense of his having never fully integrated into the host society where he worked and taught, that he was, again in his own words, the agent “straddling the borderline between cultures” (Simeoni, 1995, p. 453). It was, precisely, this position that made him nervous around the claims of extensive universality and caused him to remain eternally vigilant with respect to those “details” that ultimately cost lives and save cultures.

Daniel Simeoni’s writings revealed a constant preoccupation with the cultural origins of theorizing around translation or, more especially, the failure to reflect on the cultural embedding of these theories. Part of this concern is rooted in a nervousness around particular kinds of universalist claims that have been to the fore in discussions of translation. In this essay, we have attempted to situate Simeoni’s concerns in forms of translation thought and practice that have come to the fore in the digital age. The notions of “massive” thinking and “detailed” thinking developed by Milner have allowed us to see how Simeoni’s concerns might have a purchase on translation in an age of automation and semi-automation. In particular, the status of the “detail” has been examined to see how it might shed light on both historic and contemporary instances of translation practice. In considering different forms of universalism, we have claimed that Simeoni’s sensitivities to the function of power and context are repeatedly to the fore in the construction of different notions of the universal. In this respect, the notion of “gap” has been opposed to that of difference. The aim has been to avoid the reifying thrust of typicality which often underlies the invocation of difference and traps translation in particular forms of identitarian politics. In Simeoni’s notion of the translator as borderline agent, we find a way of restoring a sense of creative fecundity and resourcefulness to languages and cultures. This is his most enduring legacy.