The Translator’s Subjectivity

Martin Boyd

Another product of Machine Translation?

Sarita me out of the fire, because before you know it the future of humanity I did not care. She showed me the way of the spirit, he informed me that all men are equal, that the only worthy ideal is the class struggle and the victory of the proletariat, I did read Marx, Engels and Carlos Fuentes, and do everything to what? To destroy after his indiscretion.

Readers of the above paragraph could probably be excused for believing it to be the incoherent raving of a lunatic with a somewhat tenuous grasp of the English language.  Actually, it is the direct result of applying Google Translate to the opening paragraph of “La ley de Herodes”, a short story by Jorge Ibargüengoitía, one of Mexico’s greatest comic writers.

In my experience, this is a fairly typical example of what Machine Translation can do. The Internet is filled with many other, similarly bewildering examples, a number of which can be found at Careers Cabin, an online magazine, ostensibly in English, but whose editors seem to rely a little too heavily on the wonders of MT software to translate their articles.

In spite of such shortcomings, many people seem to be of the opinion that Machine Translation is getting closer every day to eliminating the need for human translators. Indeed, on numerous occasions at social gatherings, after discovering my profession, people have asked me how I feel about the fact that MT software like Google Translate will soon be making my job redundant. Obviously, these people haven’t read the articles on Careers Cabin; perhaps the only examples they’ve seen of the wonders of MT have in fact been machine-translated articles that have been extensively post-edited by human translators (an increasingly popular trend in the translation industry, “post-editing” has become an industry euphemism for “re-translating”).  My intention in this article is to answer this question about the supposed impending obsolescence of human translators; or rather, to clear up some of the misconceptions that lie behind it.

A good place to start is by clarifying how Google Translate works. Contrary to popular belief, Google Translate is not an “intelligent machine” with the ability to do the work of a human translator; rather, it is itself the work of human translators. It is, in effect, the world’s largest single translation memory, a vast database of translations previously completed by human translators. When you enter a paragraph of text into Google Translate, it searches through its vast corpus of human translations to find segments of translated text that match what you have entered, and patches them together to produce a translation. Google Translate would not exist without the millions of hours of work of human translators that have fed its database – which of course means that unless it continues to be fed by the work of human translators, it will quickly grow stagnant, as languages are constantly evolving and changing.

Even so, as the above example above shows, the highly complex process Google Translate uses to patch the fragments of human translation together rarely produce satisfactory results. Indeed, as those of us who have had the dubious privilege of working in MT “post-editing” know, machine translations are generally so fraught with semantic errors, awkward phrasings and downright gibberish that it is generally faster and easier to start a translation from scratch than to endeavour to turn a machine translation into an accurate and readable representation of its source text.  But why is this so?

The answer to this question should in fact be obvious, but the popularity of the misconception that machine translation will one day supersede human translation is the product of an underlying misconception about the nature of translation itself. This misconception is what translator and translation scholar Lawrence Venuti refers to as the “instrumentalist” conception of translation, the perspective that has dominated discourse on translation for over two thousand years now. According to this perspective, translation is a simple, direct transference of meaning from one linguistic code to another, a process of matching a word, phrase or concept in one language with its “semantic equivalent” in the other language. The problem with the instrumentalist perspective is that “semantic equivalence” is not quite as simple as we often imagine, as the semantic values of individual words and phrases are determined by their larger context. Even a Spanish phrase as deceptively simple as the connector “por otra parte” might be used in a contrastive sense similar to the English “however”, or, conversely, in an additive sense like the English “furthermore”…  or it might even be used to flag a complete change of topic (for which there is no real equivalent set phrase in English).  To understand which sense is intended requires attention to larger textual structures that is as yet beyond the ability of MT software.

Moreover, outside the text itself, beyond the structure and texture of the text, the context that shapes its meanings also includes author intention, target audience, and the social and cultural values in which the text is read and interpreted. It is hard to imagine a software program, however sophisticated, with the capacity to take all these factors into consideration.

Against the instrumentalist conception, Venuti posits what he calls the “hermeneutic” conception of translation, whereby language is viewed not as a simple representation of objective reality but as an interpretation of that reality, shaped by the social and cultural situations in which it is used. The act of translation is thus a creative act, not of extracting the meaning of the source text, but of constructing its meaning subjectively based (hopefully) on an understanding of the source culture, but also in accordance with the new context of the target culture. Indeed, it is the highly subjective nature of the act of translation that often makes it difficult for two translators to agree on the “right” translation of a particular text, as the “right” translation will depend on the translator’s subjective interpretation of the source text. Needless to say, no matter how sophisticated machine translation software may become, it is hard to envision it developing this component of subjective interpretation that is at the heart of the art of translation. As technology advances, subjectivity remains among the few human properties that machines cannot ape.

This is not to say that machine translation does not have its uses. At a talk at this year’s Multi-Languages Translation Conference in Toronto last month, Venuti remarked that he used Google Translate as a primary support for a recent translation of a book from Catalan to English, as it proved useful when searching for the mot juste in certain moments to be able to pull out at random a previous translation of a phrase from this vast database of human translations. I often make use of Google Translate in the same way, as another resource to support my work, along with dictionaries, glossaries and project translation memories. It is, quite simply, another tool for professional translators. But believing that MT software could replace translators is rather like believing that power tools could replace carpenters. Machine translation can serve to facilitate the work of professional translators; but, I would argue, it is not so much useful as dangerous in the hands of people who lack the necessary language skills and training in translation to know how to use it – in rather the same way that a power saw would be in… well, in my clumsy, untrained hands, for instance.

By way of conclusion, below is the original text of the paragraph at the beginning of this article, followed by my own attempt at translating it. My translation is also far from perfect, but I think you will agree that it at least has the advantage of being the work of a human being, capable of subjective interpretation.

Sarita me sacó del fango, porque antes de conocerla el provenir de la Humanidad me tenía sin cuidado. Ella me mostró el camino del espíritu, me hizo entender que todos los hombres somos iguales, que el único ideal digno es la lucha de clases y la victoria del proletariado; me hizo leer a Marx, a Engels y a Carlos Fuentes, ¿y todo para qué? Para destruirme después con su indiscreción.

Sarita pulled my head out of the sand, because before I met her I had no idea about where the human race was headed. She showed me the true path; she taught me that all men are equal, and that the only worthy ideal is the class struggle and the victory of the proletariat. She got me to read Marx, Engels and Carlos Fuentes. And all for what? To destroy me later with her lack of discretion.

5 thoughts on “The Translator’s Subjectivity

  1. What an excellent article. It confirms what I thought about machine translations: never rely on them for accuracy. This explains why I don’t use them when I translate texts. There is nothing better to do than to consult dictionaries, glossaries, colleagues, clients, and previous documents to find the exact term(s) of the word(s) in question. As powerful as machines such as Google Translate may be, I don’t think they will outnumber human translations in the not-too-distant future. People should discover the advantages and disadvantages of automatic translations very soon, for such translations can have serious consequences for businesses, clients, and their customers on the receiving end.

  2. Great article, Martin. I’ve actually found post-editing Google Translate not too bad for some texts; I think it depends a lot on the language pair (seems to work better for English>Spanish than the other way round) and the text type. But I agree for any writing with even a little creative flair (especially literary texts) it’s better to start from scratch than to try post-editing the mess that MT makes of them.

  3. I personally do not rely on machine translations because they are deceptive, with a lot of false cognates. The best translations are got from our use of dictionaries, glossaries, word banks and even friends.Machines cannot effectively capture the intention of an author.

  4. Pingback: Weekly favorites (Dec 24-30) | Lingua Greca Translations

Leave a Reply

Your email address will not be published. Required fields are marked *