Artificial intelligence (AI) is revolutionizing most fields, and machine learning in other areas has overcome obstacles that the translation field is still facing. For businesses and organizations that need translation services, the question is: Will Google Translate, Microsoft Translator or other machine translation (MT) tools ever be comparable to human translation?
The answer? It depends on what you’re looking for.
MT—then and now
As a tourist, you may have asked, “Hey Siri, how do I say, ‘Which train do I take to visit Pompeii?’ in Italian?” On Facebook, you never have to wonder what a comment in Japanese is saying in response to a friend’s post; just click “See Translation” below the reply, and it appears.
These are examples of relatively low-stakes communications, ideal for today’s MT to handle.
Machine translation has come a long way since debuting in the 1950s; the first systems learned how a language works by examples incorporating specific vocabulary and grammar rules, the way schoolchildren traditionally have done. Next came translation based on algorithms that compared one language’s patterns to those of another; this method was effective, but its generic nature could be inaccurate, especially when translating jargon or industry-specific terminology.
Today’s MT utilizes AI in complex layers, based on the theory that if you take the best of both older forms of MT, then add in AI, you might get a more accurate translation.
For very large documents, today’s MT is a fast, easy, and often inexpensive option.
The human touch
Not willing to risk handing over total control to machines to translate your organization’s important information? You’re not alone.
Translation of more high-stakes communication–like a medical device’s user manual, a human resources’ benefits package, legal documents, or internal corporate communications–is often where professional organizations find the “sweet spot” for qualified linguists whose training helps them find the intended meanings of words in our imperfect and dynamic languages.
“For high-quality, professional translation of content in a specific domain, human beings are very much still in evidence,” Babson College professor Tom Davenport writes in Forbes. “Human translators still translate books, articles, marketing materials, legal documents, and ‘localization’ initiatives to make product information relevant to consumers speaking different languages around the world.”
Take a frequently used word, such as “print,” for example. Is it a command? A noun? A verb? Human translators can look through a sentence, see the meaning, take the target audience into consideration, and create a more accurate translation that is more natural.
Certified translators go through extensive training to become qualified language experts. Their additional value lies in their human capacity to problem-solve, be creative and think critically, something machine programmers are far from perfecting.
If all you need is casual translation, then straight MT is a good choice, Davenport asserts. Google Translate has improved since its inception, “but it is still poorly suited to high-quality, domain-specific translation.”
When it’s obvious only MT has been used on something like a product’s packaging, the translated message often feels cheap and impersonal to consumers, who likely won’t respond well to it. Professional linguists, on the other hand, review the vocabulary and context of the statements to provide more accurate and nuanced phrasing that MT misses.
If your company has an e-commerce site, consider this research by Stripe, an online payment portal, shared by Slator.com: in looking at the top 450 European e-commerce websites, the most common error during customer checkout was the lack of translation.
“More than half (58%) of customer checkouts ‘had at least three basic errors, adding unnecessary friction for customers and complicating the checkout process.’ … (nine) in ten lost sales in Europe came from failures on the checkout page,” Slator reports. Nearly three-quarters of checkouts “did not have local language translations…(and) failed to offer the most relevant payment options for international customers.”
The privacy of your information is critical, too. Some translation services add your documents to their database for use on other clients’ projects; others do not. Knowing how your data is being maintained and used is a crucial question to pose, no matter what kind of translation project you’re tasked with executing against.
A Hybrid Future
With today’s global economy connecting nearly every industry, the need for document translation has found a home in many business and strategic plans.
Google has made great strides in its translation abilities. In fact, you can send an encrypted message to a friend in ancient Egyptian hieroglyphics using the newly released Fabricius; powered by AI, it can translate today’s phrases and emojis into 5,000-year-old symbols.
On Instagram, posts promoting a product or service can be translated by readers who can simply choose a translation language in their settings. Conveniences like these are practical and simple solutions when translating low-risk, casual communication.
But the strength of MT for business projects comes into focus when combined with the skill of a professional translator.
Say you have a 500-page proprietary document that requires translation; a linguist can apply a computer-aided translation tool to do a first pass, allowing the machine to do the heavy lifting in a lot less time. Then the linguist goes to work, reviewing the document to catch and correct inaccuracies due to technical jargon, words with multiple meanings, and grammatical errors, just to name a few.
While machines have made translations easier, faster, and even cheaper in some cases, humans are still needed to lead in the process. The ideal partnership for most translation projects involves trust and transparency. It blends “machines using the latest AI technology, translators who are comfortable with the technology, project managers who oversee the process, and clients who specify and review the content to be translated,” says Davenport.
Knowing how your words will be translated—by a machine, by a human, or by a combination of the two—can put your mind at ease, knowing your product packaging or website won’t wind up on this year’s viral list of “Top Translation Fails!”