Translation Gone Awry

The internet user base is growing broader every day, reaching more and more people outside of the US, outside of English-speaking countries, and outside of the elite English-speaking subsector of those countries.

To be competitive, companies must design robust and realistic strategies for getting their message out to customers who speak a foreign language. There are many tools available to help companies internationalize and maintain content in multiple languages. Machine translation is the most attractive option because it’s the cheapest one. But it clearly has its limits. Before you count on it to be your silver bullet, it’s worth taking a look at the limits of computer-aided translation services.

When you aim to construct a multi-lingual web strategy, it’s essential to select the right tool for the job, just like any good carpenter would. This means that you have to really think about the job you want done, and determine the tolerance for error that can be allowed.

One sort of job that can be classified as high-tolerance in terms of errors, is communicating with customers who send email messages. Taken as a basic customer service function, where speed of reply is arguably the most critical factor, machine translations can be used to understand the gist of a received message and to send an acknowledgement of its reception with a promise for follow up.

This is a strategy designed to generate goodwill and buy you time. You can use the time to get a proper translation of the customer’s message, or, more importantly, a correctly written translation of your reply. In a previous column, Messages In A Bottle: Translation Services That Work, I reviewed ways to use tools like Globalink’s Comprendre software, which automatically translates email for this specific task.

Another sort of job that has far less margin for error is the content of a corporate web site, where not only your brand image is as stake, but where errors can create legal liability. While I’ve seen companies choose to localize their sites with machine translations rather than relying on an actual human, it’s really not something I’d recommend. It’s true that machine translations are far cheaper than paying a human, but such a decision is usually penny-wise and pound-foolish.

There are two better solutions. One is to not even pretend that you care enough to translate your site for foreign visitors, but to integrate an on-site third party translation engine so prospects and customers can more easily get the gist of what you are saying or selling. Even with this strategy, I’d recommend that you pay for accurate translations of a disclaimer to go along with the “free service.”

A better solution would be to use high-end computer-aided translation services together with human translation services. It’s worth spending extra money on translations when the accuracy of your message is critical, such as when it’s a reflection of your brand or product specifications. Terms, conditions and/or claims carry a contractual obligation. Two well-known suppliers of computer-aided translations services are Lernout & Hauspie Mendez and Systran.

If you’re not convinced it’s worth the money to combine computers and humans, let’s look at an example to see why.

The following is a news item taken from the German version of MSNBC about the around-the-world balloon race which could become an international incident if the message from the Chinese isn’t understood correctly. The original reads: Eines der dr ngendsten Probleme der Weltumrunder im Ballon “Global Challenge” hat sich nun gel st: China hat den drei Ballonfahrern auf ihrer Weltumseglung nach Auskunft der britischen Botschaft in Peking am Dienstag eine berfluggenehmigung.

The computer-aided translation, from Systran, gave this version back: “One of the urgent problems of the Weltumrunder in the balloon ‘global Challenge’ solve itself now: China gave on Tuesday an overflight/flyover permission to the three balloon drivers on their Weltumseglung the message in Peking, British after information.”

A simple example like this shows some of the limits of machine translation. It’s obvious that human intervention is required when it’s critical to make sure your message is read and understood by foreign-language speaking customers and prospects without significant inaccuracies. Many of the computer-aided translation services are valuable tools. But in the international marketspace you still need a human to get the job done well.

Related reading

A push message should link the user to a meaningful page where they can learn more about your offer.
Glittering cloud with electric neurons