Google is Taking Over Translation

Google recently just revamped their Translate algorithm, and with it has been an interesting uproar and concern from the translation community. Is machine translation going to take over? What will happen to our jobs?

New vs. Old System

The original Google Translate algorithm used Phrase-Based Machine Translation (PBMT), which would translate one word or phrase at a time and then order the words grammatically correctly according to the target language.

The new system, Google Neural Machine Translation (GNMT) will use existing information on how to translate between multiple languages to learn how to translate other language pairs. For example, once GNMT learns how to translate between the languages pairs, English<>Russian and Russian<>Spanish, the system can use that knowledge to figure out how to translate the language pair, English<>Spanish. As reported by technology blog Tech2 during the pilot, “When the translation knowledge was shared, curious Google engineers checked if the [artificial intelligence] could translate between language pairs it was not explicitly trained on before. This was the first time machine based translation has successfully translated sentences using knowledge gained from training to translate other languages.”

Like Inc. says, “In other words, Google Translate’s A.I. actually created its own language, to enable it to better translate other languages.”

Compared to Human Translators

When rated by human translators, Google Neural Machine Translation (GNMT) showed a huge improvement over the phrased-based system and is nearing human levels of translation. However, with the huge media attention the new system has received, it is important companies and individuals to note the word “nearing”, as in not exact.

Translatino quality measurement comparing human, GNMT, and PBMT tranlsations
Source: Google

GNMT still makes mistakes that human translators never make. This is especially important regarding a person’s name (check out our blog about this) or words that are not commonly used. A human translator also considers the entire text when choosing phrasing, whereas GNMT only translates individual sentences.

Moving Forward

For vital medical documents, legal documents, and pharmaceutical translations, it is important that they are completely accurate, as they have a huge impact on language access and companies can be legally liable. Therefore, companies, individuals, and organizations should still seek out professional translation services for these documents.

Website localization is one area in particular that it will be hard to bypass human translators, as machine translation simply cannot adapt content to a local target culture. GNMT directly translates content without determining if the final translation is appropriate in the local context. For search engine optimization, it is vital that an experienced local individual can help ensure that the original keywords are translated into popular local keywords. Finally, compared to machine translations’ focus on words and phrases, localization experts focus on the entire project, including choosing appropriate colors, font, pictures, etc that fit within the target culture.

How Can We Adapt?

It is likely that machine translation will only get better, so as translators, it is important that we think ahead and ask ourselves how we can adapt. As several news items have stated, the use of GNMT is like the first draft of a human translation. With that in mind, GNMT might be a potential cost saver for non-technical documents and reduce the number of drafts for a human translator.

Translators also have the opportunity to turn their focus to the creative side of the business, through localization and transcreation. They no longer have to translate exact words and can have fun re-creating texts for target cultures. This is really an area where translators can show off their local knowledge. As machine translation becomes better, it may be viable for translators to market themselves within a specialty area of translation or with an added skill, such as a local marketing professionals.

While it is a scary thought that our jobs may be at risk as machine translation becomes more and more accurate, machine translation is an exciting step for language access for all. What defines the translation profession may change over time, but with a little bit of creativity and adaptability, translation will continue to be a valued profession and able to contribute positively to globalization.

Give us your thoughts about GNMT and how translators can adapt!


Leave a Reply