Science has made some great strides in developing translation software that utilizes learning algorithms and sensor technology in the last couple of years. One example of this was the release of Microsoft’s new Skype Translator, promptly followed by Google’s release of their version of a voice translation app. Both of these applications can translate spoken words into any of their supported languages.

Verbalink, a professional translation firm, decided it might be wise to measure the threat level of their newest competition. Translation can be tricky especially when you take into account local dialects and differing socioeconomic backgrounds; even words within the same language can have different subtle connotations for different people. Meaning is also affected by tone, inflection and body language.

They decided to compare the quality of translations by Adriana, one of Verbalink’s professional human translators, against the newest version of Google Translate. The criteria for quality included grammar, comprehension, cultural idioms to determine overall strengths and weaknesses. Gaby, another of our experienced translators, judged the quality of the two translations.

The results of the test showed that, while Google Translate’s version was a huge improvement over previous versions of machine translation, they aren’t ready to replace human translators for at least a few more years. If you find yourself wondering whether the results may have been skewed because the translation samples were judged by one of their experienced professional translators, here’s an example you can judge for yourself.

This is a translation of a partial phrase by Google Translate:

“…with more details about the planning of our student withdrawal of Culture…”

Now, here’s Adriana’s translation of the same phrase:

“…in more detail about the planning of our student cultural retreat…”.

This example is a good illustration of the inability of machine translation to take context into consideration when determining word choice. While not even human translators can use body language cues through voice or text, they still do a better job of providing similar meanings for idiomatic expressions as well as providing cultural context.

Artificial intelligence has come a long way, and it can be handy to have a translator that fits in your pocket. These apps are great for getting the gist or for communicating under casual circumstances. When you need a precise translation, in which cultural and emotional content can be just as important as verbal content, there’s still no substitute for human experience.

You can see the complete results of the comparison below.

human translation vs machine translation

This post was written by Ashley Harris on behalf of 7labs.