Can we trust Machine Translation?

As online translation machines become better, it is tempting for many to use these instead of human translators or professional translation services. However, Anne Merritt warns us that machine translation must be used with caution, especially for longer texts.

In her article on The Telegraph, Merrit draws attention to the pitfalls of online translation devices.

Machine translations (a.k.a. translation software or computer translation) that were once seen as science-fiction gadgets, have not only become a reality but their accuracy is rapidly increasing. Social media sites such as Facebook and Twitter have translation options and machine translators such as Bing are in the wake of the biggest player on the field, Google Translate.

Both Bing and Google Translate use “Statistical Machine Translation”, or SMT. This is a type of computational linguistics that uses large databases of text to trace translation patterns and draw parallels between languages. These translation machines thus do not carry out word-by-word translations, but use word combinations and common patterns to produce a translation. Their software is improving constantly through the input of new data. For example, Google Translate encourages people to improve the translations they produce which they then use to improve their SMT.

This might seem like a great translation method, but Merritt asks herself just how reliable this kind of software is.

According to her, there are a number of common flaws that occur in existing translation software. Everyday phrases, such as ‘get over it,’ are not always detected. In addition, the syntax of longer texts is often too difficult for the software to handle.  It is for example unlikely that a translation device can correctly translate a sentence such as “I would have gone swimming yesterday, but the weather was so lovely that I took a walk instead.”

Despite these flaws, Meritt believes these translation tool “do an impressive job,” especially with short texts such as Tweets or recipes. For longer texts that involve more details, such as academic texts, Meritt however discourages the use of machine translation.

“As an ESL teacher, I witness a handful of student essays each semester that have been run through some kind of translator,” Meritt says. “I think every language teacher will agree when I say that yes, we can always tell. Often, hefty portions of the text are just incomprehensible.”

And illegibility isn’t even the biggest problem according to Meritt. Faulty translations can also mislead or, even worse, offend readers. A great example of this is a slogan mistranslation by Pepsi. The Chinese translation of their slogan “Come alive” was “We bring your ancestors back from the grave.” This might seem a funny story, but Merrit believes it serves as a cautionary tale for people who wish to use translation at work.

Translation involves more than just a word-by-word representation of a text; translators also have to take double meanings, cultural subtleties and slang into account. That’s why when your document or text is important that you use a professional translator or an accredited agency.