The Alchemist's Lair
Thoughts on writing, language, life as a freelancer,
and what comes to mind.
and what comes to mind.
Londa Schiebinger, head of Stanford’s Gendered Innovations project, recently published a case study on machine translation and gender neutrality. The research showed a tendency of machine translation tools to render gender-neutral English words as male words when translating into languages using gender-specific structures (such as Italian, German or Spanish to name but a few). However, when translating words associated with cultural stereotypes (such as teacher or nurse), these words were translated as feminine.
As a professional translator myself, I have witnessed a growing trend of relying on machine translation tools both for personal and business use. Crisis is hitting hard in Europe and abroad, and businesses try to cut costs whenever possible. This is completely understandable – we all keep eyes wide open on budgets nowadays. However, culturally biased translations and accidental racism are not the only instances where free turns out to be pretty expensive and harmful.
Last year a prospect asked me for a quotation for his e-commerce website translation. My price was out of their budget, so they looked for a more convenient solution. A few months later, however, the company got back to me as they received dozens of complaints about their new Italian website. Apparently, customers actually preferred to shop on their English website, as the Italian version was barely understandable.
As a result, all their investments in AdWords campaigns and other marketing actions to promote their new website didn’t result in the expected sales volume. A complete waste of money, and the reputation of the company amongst their Italian business partners was at stake. It turned out that the free machine translation tool they used did more harm than good to their business – and errors were not just in “the tough stuff”. As an example, “3 days left” for their limited offer had been translated as “3 giorni a sinistra” (3 days on the left). A mistake no skilled professional would have ever made.
If you ask a human translator what does a word mean, the answer would probably be it depends on the context.
Think of words like get, play or set. They all come with different meanings (and some of them with literally dozens of meanings) depending on the context of use. Some of them are well known, some others are more obscure and old-fashioned. However, being old-fashioned does not prevent anyone from using a specific word with that meaning.
As professionals, we are trained (and paid) to spot all of those sneaky instances where words can be confused with homonyms, or used in unusual or figurative ways, or maybe just misspelled by the author of the original text. To do so, we rely on a mix of academic background, experience and expertise. At some point in our career we all made mistakes, and learnt from them. The human mind works based on dynamic processes allowing us to do so. Also, when we struggle in grasping the pitch of a sentence, we can just pick up the phone and ask the customer for further clarification.
Translation machine tools cannot do that yet. Machine-operated translations are based on statistical analysis and tied to the search database used by the tool, which operates based on rigid algorithms. Said algorithms take into account quantifiable aspects of language (such as spelling or grammar structure), but cannot analyse the context of use of the word, nor ask for further clarification if needed. As a result, accuracy in translation is not guaranteed.
Accuracy, however, is not always necessary.
If you want to read a blog post in a language other than yours because you’re interested in the topic, machine translation might provide you with a general understanding of what’s being discussed. But when it comes to business, you really want to make sure you get it right the first time to avoid losing sales opportunities and accidental offence to your customers or readers. And the only way to do so is to rely on the skills and expertise of a professional.
As you see, it all depends on the context after all.
This blog post was originally written as an Op-Ed assignment for Duke University's English Composition I: Achieving Expertise MOOC offered on Coursera. The examples reported are real, and opinions are my own.