Translation Technology 101:
How It’s Changing the Business and How We Use It

Part I: What You Should Know about Machine Translation

With all of the recent technological advances in machine (or computer) translation, we realize it’s easy to wonder whether the folks at Avantpage are just plugging your copy into some software before we send it back to you. The short answer is no, we definitely don’t do that.

We’ve created this special report to help you learn a little more about the way machine translation works, to explain the way technology is helping us at Avantpage work more effectively, and, finally, to answer any questions you may have about “translation tech.”

An Extremely Brief History

The process of using computers to translate between languages began in the 1950s. After a successful experiment at Georgetown University, in which Russian sentences were translated into English, researchers were so encouraged that they claimed they would have a definitive solution within 3 to 5 years. This, of course, proved to be unrealistic due to the enormous challenges and demand on computing power involved. Funding was decreased in the 1960s after reports showed that little progress was being made.

In the late 1980s, as computers rapidly became more powerful and less expensive, machine translation once again became a feasible research subject.

Two decades later, according to LISA (The Localization Industry Standards Association), more words are translated each year by machine translation than by human translation. The trend is only expected to continue.

How Do the Machine Translation Programs Work?

Currently, machine translation programs can be grouped into three main categories:

1. Rule-based translation—Produced using existing grammar rules and dictionary entries.

2. Statistical translation—Uses data from a large amount of existing bilingual material, known as a corpus, to translate commonly used words and phrases. Examples of multilingual data include the records of the European Parliament, which is produced in 23 languages. Google Translate has been using this type of software since 2007.

3. Hybrid machine translation—Ideally uses the best features of both statistical and rule-based translation. SYSTRAN, which powers Babel Fish, claims that it introduced the first hybrid translation engine in 2009.

Google Translate: A Game-Changer?

Google's free Translate application is now able to handle 52 modern languages. This is extremely impressive, considering the fact that it has taken Google only a short time to do this compared to other machine translation companies. One of the reasons is that they have moved beyond the traditional data such as the United Nations transcripts and used texts taken from all over the Internet, as well as data from the Google Books project.

Instead of using a billion words to model the English language, they have used several hundred billion words. Franz Och, Google’s leading translation scientist, told The New York Times, “The models become better and better the more text you process.”¹

The result is a translation engine that is excellent if you need to get a fast, rough translation (such as reading a news article from another country) to simply understand perhaps the gist of a document. However, a machine-translated document is risky to use when you want to effectively and accurately communicate to a specific audience.  In addition, this document will likely lack cultural sensitivity.

Better Machine Translation=Better Human Translations

Even though machine translation is very useful, it still has its shortcomings. The problem with using pure machine translation, where a user inputs one document and gets a translated one back, is that the end result is extremely rough and your audience will spot errors instantly.

That's where human translation experts can take over. Machine translation is best used as part of a quality-focused process in which humans control and monitor the final translation product.

So will the computers be taking over the translation world? The answer: Not any time soon. But the recent innovations by Google and others are bound to have a lasting impact on the industry.

Is Avantpage Using Machine Translation?

In a word, no. We use technology as tools within the process, but we don’t use any of the methods listed above.
There is no part of our process where we submit text into computer program and then get a translated document back. We use humans to translate—human linguistic experts who are native speakers, know industry terms, and know local dialects. Because we treat every document that we produce with the utmost care, we choose not to use machine translation tools at this time.

In Part 2 of our report, we’ll discuss the ways we DO use technology to make the process more efficient.

*********

About Avantpage
Avantpage has more than 13 years of proven excellence in linguistic services. We offer language services such as translation, interpretation, localization, and multilingual desktop publishing designed to perform efficiently, consistently, and accurately. At Avantpage, we use state-of-the-art translation tools in order to improve efficiency and overall quality for our clients.

To learn more about the way we’re using technology for better translations, or to discuss your translation needs, call This e-mail address is being protected from spambots. You need JavaScript enabled to view it , VP of Business Development, at 530-750-2040, x7. We believe that business is personal and our consultative style ensures that it is.

References:
1. Helft, Miguel. “Google’s Computing Power Refines Translation Tool.” The New York Times. March 8, 2010.