“Translation is clearly too important a task to be left to machines.”

Julian barnes, LRB Vol. 32 No. 22 , 18 November 2010

If a business has gone to the trouble and expense of securing itself a first-rate copywriter, the chances are they’re not going to simply feed the resulting text through Google translate or perhaps even Deepl to generate the foreign language versions. (All the same, doing just that can provide a rich vein of awkwardness and hilarity at the expense of an algorithm, a harmless enough pastime as the machine really couldn’t care less. I shall be looking at some examples of this in a later post.)

Whether implicitly or explicitly, texts which intend to persuade, and it could be argued that all texts have this aim, will incorporate certain aspects of Aristotle’s rhetorical triangle from rhetorical analysis, or ideally all of them to a varying degree:

ARISTOTLE’S RHETORICAL TRIANGLE:

Ethos: establishing personal credentials, reputation,

gaining audience’s trust, establishing tone and style,

confidence in delivery, personal branding

ETHOS

LOGOS PATHOS

Arguing based on reason, facts, Inspiring an emotional

case studies. Proof: Getting response. Appealing to

your audience to think. audience’s values. Storytelling.

Inspirational quotes.

Vivid language.

In her recent LinkedIn post on machine translation and MT post-editing, language expert Ingrida Tatolyte said:

“… when we come to nuances, and whether the same phrase would actually be used in target language naturally, what connotations it would have, would it have the same emotional charge or not, those are missed when post-editing, but corrected in edition of human translation.”

I would go further to say that when one human translator corrects another human’s translation the creative flow is still distorted and the freshness and the impact of the language can be lost, because the proof reader is not experiencing the text directly and thus applying their own creative strategies but is viewing it through the filter of the original translator’s viewpoint and limitations. This is particularly true if the original translator is not very highly skilled: in a very competitive market it is not uncommon as a skilled translator to be to asked to “correct” a translation by an inexperienced and thus very cheap translator (or even a text written in English by a non-native speaker, which is another kettle of fish). The ethos and pathos in the text thus become weakened or are poorly reflected in the final version, particularly if the text is very long (I’m talking 45,000 words or more) and the proof reader becomes gradually and subtly influenced by the original translator’s (lack of) style. In effect, although the bare facts are still stated in the target language, in other words the logos has largely remained intact, the overall message has essentially been diluted or lost. This dynamic is far more pronounced when a skilled translator is asked to post edit a machine translation. That said, a proofread text is necessarily always better than an uncorrected one, as four eyes are better than two.

Ingrid Tatolyte went on to say:

“…logos vs ethos/pathos will always depend on the situation and context, much broader context than machine can cover, and will remain a more sensitive aspect, as machine can easily manipulate such things given algorithms, etc. I think there are areas that must remain human responsibility.”

Ingrida Tatolyte, Language Professional / Lecturer of Practical Rhetoric / Researcher on LinkedIn

Having said that translation was too important a task to be left to machines, Julian Barnes went on to examine what sort of human it should be given to by comparing the various English translations of Flaubert’s novel Madame Bovary which have been published over the past 150 years. This delightful excursion into the world of literary translation is well worth a detailed look, as it reveals a great deal about how to go about choosing a translator. If you don’t have time to read the whole article in the LRB, I’ll be quoting from it in my next post.

#machine_translation #neural_machine_translation #human_translation #post-editing #Aristotle_rhetorical_triangle