Home Get a QuoteBlog
1-800-320-1020 (North America)+44-20-3393-8385 (Europe)+972-3-5338633 (Middle East)

Is Machine Translation Good Enough?

Is Machine Translation Good EnoughYou might have heard of machine translation (MT) and thought it sounded like an interesting and cost-effective solution. You just throw your content into a machine and it’ll be done quickly and cheaply, right? But what about the quality of the translations? Let’s take a closer look at machine translation.

So What Exactly is Machine Translation?

Machine translation is software that translates source content from one language into one or more other languages. Google Translate is an example of MT and I think we’ve all probably tried it with mixed results. There are two types of MT technologies and both approaches have their strengths and limitations:

  • Rules-based: this type of translation applies standard language rules like grammar, meaning and syntax to each source segment for each language. Rules-based MT can be effective in cases where there’s high-quality content that’s pretty straightforward and it isn’t specialized.
  • Statistical: this type of translation uses existing translation tools like a translation memory (TM) and a terminology database to analyze the source language and identify the closest translation. In this case, the MT is “trained” because it’s fed existing translated content, and over time, it becomes more trained. Statistical MT can be effective when there’s a lot of existing content or the content is specialized.

What’s important to understand is that machine translation should not replace professional human translators. A machine is just that – a machine. No matter how much content you feed into a machine translation, the software just won’t get all the nuances and culture of a language. Only a human can do that.

What you need to ask yourself is: as a global company, would you tolerate mistakes in your English content? If not, then machine translation should not be used. At the very least, it should not be used without human translators in the post-editing process.

Improving the Quality of Machine Translations

As I mentioned, machine translation is just software. It’s pretty well known that the quality of machine translation is still pretty low and not comparable to professional, human, translated content. You can train the machine to get to know your content by adding the translation memory and the glossary terms, but it’s still not good enough to stand alone.

What can help improve the quality is implementing a post-translation editing process, which involves having a human translator review and correct the translations. In this scenario, instead of starting from scratch, a translator starts with a rough translation. Of course, this will add cost and time to a project, and at some point, you have to decide if machine translation is worth it or if it’s better to just use human translations from the start.

When is Machine Translation Good Enough?

For the bulk of your business content that your customers will read and use, such as website copy, marketing brochures and technical instructions, you want to have the highest quality content possible. In this case, you must use human, professionally-trained linguists. In cases where the quality may not matter quite as much, such as for customer-generated content like a customer support forum or product reviews, and you need quick turnaround times, machine translations might be good enough. Also, if you reuse a lot of your content and you have a large amount of it, then training the software may reduce the costs and save time on the post-translation editing process.

For example, Jose Luis Bonilla Sánchez, Machine Translation Language Specialist Supervisor at eBay, uses machine translation to manage the sheer volume of content that they have to manage (12,000+ product categories and hundreds of millions of words that are translated monthly). The more transactions that happen, the better it is for the buyers, sellers and businesses, so time is of the essence at eBay. Jose has several quality checks built in throughout his translation process, including manual translation reviews. He believes strongly in the value these translation reviews provide. Jose was one of three guests in our webinar on localization quality. Read more about Jose’s process in our blog post and I hope you’ll find some time to view the webinar at your convenience:

A New Type of MT: Google’s Neural Machine Translation

Google has been working on a new type of machine translation called neural machine translation (NMT) that they’re saying is immensely better than Google Translate. It hasn’t been released to production yet as they’re working through issues, so it’ll be interesting to watch what happens with it. Read more about what experts are saying in this article.

Conclusion

Overall, you need to develop a content management and translation strategy that makes the most sense for your specific business. Our team of globalization experts will gladly answer any questions you may have and we invite you to reach out to the office nearest you.

About the author:

Shy Avni

Shy Avni is a serial entrepreneur who co-founded Net-Translators in 2002. Under his leadership Net-Translators has become a leading localization company, operating offices across the world focusing on providing services to the technology industry.

Contact Us