ULG’s Language Solutions Blog

When Machine Translation Works, and When It Doesn’t




If you haven’t been living under a rock for the past year, you’ll know that Machine Translation (MT) and its applications continue to gain traction as one of the hottest topics in the language industry.

The ongoing debate on whether this new technology will replace human translators in the near future has been filling up our social media feeds and capturing our attention on industry blogs.

So, what does the future of MT really hold? For now, it’s hard to say definitively what the technology will look like in 10 years. But, we do know that for the time being, it’s impossible to get by only using MT for all of your translation needs. The technology has not yet proven itself as adept at deciphering linguistic nuance as human translators.

But that’s not to say that the idea of using Machine Translation should be thrown out the window. It has great benefits for those needing translations done fast and cheap. But, it’s important to know when to utilize MT and when to stick with the human touch.

Saving Time and Money

Maybe MT’s biggest benefit is its ability to save companies money. Using Machine Translation as a first pass for large volumes of source documents or to get the “gist” of foreign language content helps eliminate the cost and labor needed for human translation.

MT can translate reams of documents in half the time a human linguist can, making it a great tool for projects that require tight turnaround times. The technology is also beneficial when only a summary translation is needed.

Using Language Identification (LI) and keyword searches, for example, can be a great way to examine a massive amount of legal documents to see what needs to be translated and what documents might not pertain to a case.

With that said, it’s important to note that quality overrides quantity. Although MT might be a quick alternative to good old fashioned human translation, it should be noted that quality suffers using even the most robust MT systems.

Neural networks have proven to be an impressive feat in the MT field, but current systems still don’t have the ability to decipher nuance or cultural relevance that humans do.

Quality Issues

So, even though MT is a viable option for certain projects, it’s definitely not for others. Any translation that needs to be 100 percent accurate and will be used externally should not be translated using only an MT system.

MT usually works best for documents like technical manuals or other documents that contain simple, repetitive text. In contrast, marketing materials using a good amount of marcom flair likely won’t come out of an MT system accurately. Any documents that contain complicated language don’t do well with MT.

And, if nuanced documents are machine translated, it’s likely the results will be of such quality it will be hard for even the best post editor to refine. That means that sometimes, MT can add time to a workflow if it’s not used correctly.

The Payoff

As a general rule, MT should be used with documents that don’t require complicated translations and aren’t expected to be 100 percent accurate. Using MT to translate a large amount of documents quickly, or to get the “gist” of a source file, is recommended.

MT should not be used when accuracy is key, or when the source language is filled with nuance and difficult terms.

To recap, Machine Translation can be helpful when:

  • Translating internal documents (emails, company communications, international communication)
  • Translating a large number of documents (Using LI for legal documents, very large projects)
  • Translating technical, simple language documents (product manuals, labels)

Machine Translation can be your worst enemy when translating:

  • External communications
  • Documents that require 100 percent accuracy
  • Source documents with complicated, nuanced language



Google+ Facebook LinkedIn Twitter Instagram


Strategy I Technologies