Why You Should Be Concerned About Machine Translation

The popular press likes to report on the speed and ease of use of Machine translation (MT). Not to mention, it is often completely free! While it sounds awfully attractive, the reality is that the technology isn’t accurate enough for most corporate or high-importance translation projects. Machine translation is especially not appropriate when the language service provider doesn’t disclose that MT was used in place of human translation. That is an inappropriate tactic and one that isn’t even allowed by the translation industry ISO standard (17100).

Why does it matter where your translation has been? Or better stated, why you should care who worked on your translation.

The key consideration for the translation buyer is what is the risk of a mistake in the translated content. Obviously, for clients producing medical devices, technical instructions for heavy equipment, and legal content, the risks are high. A mistake in translation can be costly. The popular argument is that translation firms using machine translation will typically have a human editor review the content. So it should be just fine! Well, that sounds great but here are the key issues with that workflow.

Experienced and higher-paid translators typically won’t work on machine translation

Our industry has an interesting way of compensating translators. Most tasks like editing and translation are paid for by the word. Basically, these per-word rates are meant to approximate an hourly rate. I hesitate to make broad sweeping statements, but for the most part, the net hourly rate of translation is higher paying than the rate for MT editing. This means that the population of possible translators is limited to translators with less experience.

The other issue is that MT editing projects are typically crowdsourced to the lowest bidder. Many of the larger firms use this type of system to facilitate large projects efficiently. This certainly makes the recruiting and staffing aspects work very smoothly but does this affect quality? Most definitely. I can’t imagine a recipe for success when quality translation includes blindly adding translators and editors to a project based on who bids the lowest amount and raises their hand first in a crowdsourced initiative. This is a race to the bottom.

Editing machine translation output is monotonous and difficult

Editing machine translation output can be incredibly difficult for extended periods of time. The work can be very monotonous. The challenge is that editor’s fatigue can set in due to poor quality. Over time, segments that should be modified are simply accepted. In my opinion, the fact that MT output is edited by a human is simply a false sense of security.

I am a strong believer in using MT for gisting. In other words, just use the machine translated content to get a broad understanding of the source content with no attempt at improving the original output. There are many instances where a broad understanding is all that is required. MT is perfect for this use case. Adding human editing is of limited value to the end-user. Of course, if your goal is to improve an MT workflow then feedback is critical. The issue is that most corporations or entities that need highly precise translations are not willing to be part of the experiment; nor should they be. They want good results now.

Know what you are getting

As I have said in previous blog posts, there are certainly proper use cases for machine translation. If you want to learn more about machine translation, check out this episode of our podcast.  The bottom line is that all translation buyers should be aware of how their translation is handled. Understand the workflow and understand if the process is using humans or machines.

If you are a translation buyer and you want to reduce costs, I am a firm believer that more energy should be spent on writing consistent and controlled English (source content), establishing a comprehensive glossary, and setting a strategy for proper management of the translation memory. These steps will provide you with high-quality translation and lower total translation costs.

To learn more about best practices in the translation industry, download our “Top ten things to consider for your next translation project” e-book here.



Related Blog Articles

4 Ways To Translate Survey Responses
Read article ›
Machine Translation Vs. Qualified Human Translation
Read article ›
Can Machine Language Translation Ever Beat Human Translation?
Read article ›
machine translation written on whiteboard
Post-Editing Machine Translation Matters
Read article ›