Human translation vs. machine translation: who wins?
May 28, 2019
Never miss an articleSign up for weekly content
I, Robot, Blade Runner, WALL-E. Three examples in which machines replace humans, for both manual labor and original thought. These works of fiction touch on an interesting philosophy: what if robots don’t need us anymore?
You could argue that rational thought will always divide humans and machines. Our ability to reason makes us “exceptional” beings. But how do you know you, yourself aren’t a computer? What if the thoughts you think have all been programmed?
What if computers were programmed to think, rationalize, and create the way we do, based on a mixture of chance and pros-and-cons? Don’t they do that already?
We create computers and machines to make our lives easier, to keep on the cutting edge competitively, and to fit in. But most importantly, they’re cost and time effective. Although machines have their own unique costs associated, they don’t need health coverage, liability, payroll, vacation time, etc. And when competitors are investing in cheaper, more efficient means for the same job, what option do you have but to follow suit?
Translations for different purposes
We’ve all experienced translation fails, whether with Google Translate or even WeChat’s translation tool. These issues make machine translation seem unreliable. So how, then, can companies like Systran and SDL claim their products will solve world peace, or that they can translate content significantly faster with machine translation over human translation?
These claims are nothing new. In fact, Systran has been around since its development in 1968 with beautiful intentions to collapse communication barriers that fuel conflicts. It’s been used in the US Airforce and Nasa to break down language barriers and translate instructions and documents. It’s available for business and personal use and is even behind the development of Babel Fish, the first free online translation tool.
There’s a lot of software options available that advertise the ability to autonomously translate huge amounts of text into various languages. Additionally, many language learning apps and machine translation programs can help users understand common expressions and an “emergency toolkit” of words.
Want to learn more? Click here for more business resources.
Google Translate works to translate documents and web pages, which can help with navigating a site and obtaining the general gist of information.
But, can they be relied on? Try it for yourself.
And aside from language learning apps, can machine translation be as effective and reliable as human translation? Will they soon make human translation obsolete?
Lost in translation
The history of machine translation includes all the machine translation struggles you’d expect to see sustained over time.
Generally, machine translation lacks an accurate depiction of context and contextual rephrasing that can make translated sentences less ambiguous and awkward.
When trying to replicate the same meaning from one language, word-for-word translations are inaccurate. Certain meanings are often “lost in translation”. It takes some reasoning to understand certain context.
English, especially, uses a lot of figurative, symbolic, or indirect language. And that’s not the worst of it. We must consider the uniqueness and nuances of all languages that native speakers and human translators can understand.
Most idioms (sayings) can’t literally translate to another language. The content doesn’t usually reflect the context; idioms don’t directly say what they mean. They’re based in a certain context, usually with deeply-rooted and forgotten about origins. They’re familiarized by common use.
But machine translators are limited to certain rules which can’t be applied to all languages. Languages don’t all follow the same word order. Arabic is a language that suffers particularly with most translators, including Systran and Google.
Imagine languages like Mandarin in which the combination of certain symbols can completely alter the meaning of a sentence, almost like prefixes and suffixes in English.
A real-time translator might identify the individual characters but can’t convey the meaning based on those combinations.
Additionally, English has many homonyms (words that are spelled and pronounced the same but have different meanings, such as, arm, watch, and cool). They have multiple meanings and parts of speech. Words like these become hard to translate because of the possible meanings and synonyms they have.
Machine translation: practical but unreliable
What about conversational, dialectical, and slang use? When traveling, people sometimes rely on apps like Google Translate to read street signs. These signs, however, are very simplified and therefore their meanings are very hard to distinguish. Whatever context and rules the app would rely on are missing.
However, Google Translate is super helpful when you’re on the go. I visited China in 2018 and relied on it and similar apps a bit too much. They helped with ordering from menus and forming sentences word-for-word.
Everything you need to do business in China, all in one place.
But sometimes even menu items and street signs were veiled by an indecipherable context, making certain foods sound…concerning. Context could mean the difference between “jerk chicken” and “rude and obnoxious coward”.
What helped most of all were the friends I made that could translate for me. Real, living human translators.
Machine translation is convenient, easily accessible, faster, and generally much less expensive than human translation. But these strengths are limited in terms of reliability. Sometimes the worst mistake a business can make when translating something is simply not knowing they’ve made an error. You might as well get a Mandarin symbol tattooed onto you without knowing the meaning.
Why should businesses invest in accurate translations?
You might have noticed that even advanced automation technology, like Grammarly and other spell-check systems, have their flaws. Grammarly had the ability to outsmart even the best writers and even add in a human touch. Although such automated addons smooth over the awkward bumps in texts, there are still many mistakes surrounding context and word order.
It’s important to be critical of suggestions made by such programs and to use your own judgment. And yes, this relates to translations as well.
Even if a business chooses to use a machine translation, it’s bold to trust the program alone. Businesses need to proofread translated documents before they publish. Who better for the job than a human translator – someone familiar with both (or more) languages?
Is that plausible all the time? No, not for huge amounts of text in which grammar and exactness is crucial. Think about how Systran worked with Xerox to distribute manuals internationally. Have you ever read a manual with a perfect translation?
Are wire transfers safe? Download our white paper to find out.
Machine translation is a great idea for practical use. But it’s not ideal for businesses with public or crucial information.
Companies invest a lot in their writers, whether for blog content, PR, white papers, even internal communications. Businesses invest in their communication because it’s what connects them to their markets and their whole teams. And, yes, translation is a big investment; it comes back in big ways.
Good translation seems natural and won’t really stand out. That’s a good thing. But poorly translated texts show a lack of responsibility by businesses. If a company publishes something that’s poorly translated, they lose credibility. Additionally, it shows a lack of compassion for the audience, as they will feel the business does not represent them.
The effectiveness of machine translation has been developing for many decades. Only time will tell how human translation and translation agencies will survive against future software.
Simplify your business payments with Veem.