Outsourcing + Machine Translation = Failsauce

If you haven’t already, register a username for yourself so you can discuss this topic in our Offshoring vs Onshoring forum.

Do you know what’s worse than using a machine translator to collaborate with a non-English speaking service provider? Believing that using a machine translator to collaborate with a non-English speaking service provider is okay. You wouldn’t know otherwise if you let the “Internet” tell you.

It appears that some people are taking the global advantage too far, and are recommending English-to-non-English outsourcing through the likes of Babelfish or Google Translator as if they were adequate and appropriate outsourcing tools. The fact is, outsourcing is difficult enough in one language — worsened when parties are restricted to text communication.

Via text, emotions are easily misunderstood for example, humor is often misinterpreted, and commonly used phrases in one area of the country are easily taken out of context in another. To throw in another variable (i.e. a foreign language) is unquestionably the wrong thing to do, unless you’re 100% well versed in that variable.

To Understand, Consider the Following

Language Translation Hire Top U.S.A. Human Language Translation coders w/ ExpertRating status and more.

justoutsourcing

Let’s say you, an English speaking person, hire a non-English speaking French programmer over the Internet. The only thing the two of you have in common is a connection to artificially intelligent translators. One of the most important tasks you need completed at the moment is a slick looking, but functional interface. When you plug that need into Babelfish in hopes of communicating it to this poor programmer, this is what you end up with:

What you type into Babelfish:

“I need you to build me a program with a Web 2.0 looking interface and a fully functional back-end database.”

What Babelfish returns and what you email your programmer:

“J’ai besoin de vous pour me construire un programme avec une interface de regard du Web 2.0 et une base de données principale fonctionnelle d’a entièrement -.”

What the French programmer may read:

“J’ need you to build me a program with an interface of glance of Web 2.0 and one principal database functional d’ entirely has -.”

Already we have a situation where (1) the requested program’s interface could be slightly Web 2.0-ish (as interpreted by ‘glance’) instead of wholly Web 2.0-ish, and where (2) the entire program may contain more than one database (as interpreted by ‘principal’). Babelfish isn’t solely to blame, however. You get the same problem using Google Translator.

What Google Translator returns and what you email your programmer:

“J’ai besoin de vous pour me construire un programme avec uneinterface Web 2.0 à la recherche et une base de donnéesentièrement fonctionnel back-end.”

What the French programmer may read:

“I need you to build me a program with a Web 2.0 interface to search and a fully functional database backend.”

Here, the French programmer may think you want an interface that can be searched, or that the sole purpose of the interface is to perform a search. S/he may also have no clue as to what a “database backend” is. Is it the same as a back-end database? And when did ‘search’ become a requirement in the first place?

Machine Translators Aren’t Ready

Current events can be found in our Translation News section.

No one disputes that machine translators are cool, fun, or helpful in some way. Thing is, they aren’t always appropriate. Machine translators just aren’t ready for the exact needs of outsourcing. They, in fact, introduce risks you don’t need: misinterpretations, miscommunications, and of course, consequential delays.

Save these translators for when you only need to get or communicate the gist of something. Use a human translator (or hire locally) when you need to outsource!

External Resources:

1. Machine translation or Faulkner?
2. Push International – Professional Translation Services