Does Machine Translation Make Sense?

September 3, 2019 |

Machine translation is the translation of text by a computer, without human involvement. You may also hear machine translation referred to as automated translation, instant translation or automatic translation.

There are three types of machine translation systems.

The first is rules-based. A rules-based systems combines language and grammar rules with dictionaries for common words. Specialist dictionaries focus on certain industries or disciplines. Rules-based systems typically deliver consistent translations with accurate terminology when trained with specialist dictionaries.

The second is statistical. A statistical system doesn’t know language rules. Instead it “learns” to translate by examining large amounts of data for each language pair. This system can be customized for specific industries using additional data relevant to the appropriate sector. Statistical systems tend to deliver translations that are more fluent-sounding, but less consistent.

The third is Neural Machine Translation (NMT). This is a newer approach that makes machines learn to translate through several processing devices that have been modeled on the brain. The approach has become increasingly popular amongst MT researchers and developers, because NMT systems have improved tremendously in many language pairs when compared to the phrase-based statistical approach.

Machine based translation may have its place. For example, Google Translate can translate into and from more than 100 languages.  But the results are not always useful. There are all kinds of stories of its delivery of funny and/or embarrassing translations.

Dogs are Dooming the World

One of the most well-known is the Google Translate glitch that translated the word “dog” typed 20 times from Yoruba to English as “Doomsday Clock is three minutes at twelve We are experiencing characters and a dramatic developments in the world, which indicate that we are increasingly approaching the end times and Jesus’ return.”

What?!

Belts, Pants, Whatever

The Boao Forum is a high-level forum where government, business, and academic leaders share their visions on important issues in Asia as well as the rest of the world. Last year they used AI-powered machine translations for both real-time interpretation and transcripts.

Unfortunately, the machine translation wasn’t completely capable of handling the intricacies of interpreting how actual humans speak. Add in speakers speaking English with various accents and Mandarin with various accents and errors are multiplied.

The resulting output frequently didn’t make any sense at all. At one point, China’s “One Belt One Road” initiative was translated as “One Road, One Waistband!”

Don’t Have a Cow!

When Israel’s Netta Barzilai won the Eurovision Song Contest in Lisbon, Prime Minister Benjamin Netanyahu tweeted congratulations. What he said was, “Netta, you are a real darling. You have brought much respect for the State of Israel.” But what Microsoft Translate turned it into was “Netta, you are a real cow.” Yeah, that’s not embarrassing!

Good Drugs

Chris_in_Georgia wanted to compliment his Italian host’s amazing meal, but when he used Google Translate for his praise, he ended up getting the word “amazing” translated as “stupefacente,” which refers to heroin.

It’s Some Holiday

Then there was the translation of Lá fhéile pádraig sona duit! from Irish to English as Happy Valentine’s Day! The correct translation is Happy St. Patrick’s Day!

The bottom line is regardless of how easy and inexpensive machine translation may seem, there’s a good reason why skilled human interpreters and translators remain relevant and important. After the Boao Forum fiasco, researcher Lily Wang was quoted in the South China Post as saying, “The difficulty is that people talk with emotions and context, but technology can’t read people’s feelings. Most programmers who develop these AI interpreters usually do not have a good understanding of linguistics.”

Alpha Omega Translations provides translation services for more than 210 languages. Contact us to ensure your translations make sense, are both fluent-sounding AND consistent, and are customized for your industry.

Category: Translation Services

Skip to content