Google Translate and other free online translation tools can be great for instant, informal translation. When expectations are properly set, particularly for low-value text, unedited machine translation can be quite useful. However, when a user overestimates machine translation capabilities, the results can be confusing at best.
When one online machine translation tool apparently mistranslated a common Chinese word as “Wikipedia,” Chinese menus began popping up everywhere with English translations for menu items like “stir-fried Wikipedia" and “barbecued Congo eel with Wikipedia and fermented bean curd.” Though odd, the error is relatively harmless. However, when the text has important implications in law, finance or marketing, the results can be terribly costly.
Potential customers reading marketing materials may get the gist of a translation, but successful marketing text usually needs to convey more than just a general idea. Wayne Bourland, a senior manager on Dell’s global localization team, noted, in a recent usability study conducted in Germany, Dell observed that… "buyers who needed to form an emotional connection as part of the purchasing process were both distracted and disappointed by translation errors.”
When a Moscow-based marketing firm asked my company to review some previously translated marketing Web pages, we had to tell the company it paid a lot of money for what was actually a very crude machine translation. If this marketing company and its clients had expected machine translation, the news would have been acceptable. Unfortunately, the firm and its customers were expecting high-quality translations that captured the nuances of the original marketing text. The need to pay for a complete retranslation by professional human translators was a bitter pill to swallow.
In a 2010 legal mishap, “a Russian trucker in (the Netherlands) involved in a bar brawl was released because the (court) summons he received was poorly translated from Dutch into Russian using Google Translate,” reported the Dutch-English news blog 24oranges. Instead of reading, “you are to appear in court on 3 August 2010,” as it should have, the summons said something more like “you have to avoid being in court on 3 August 2010.”
This column has previously mentioned many other incidents resulting from improper use of machine translation. A Chinese restaurant sign displayed the words “Translate Server Error” above its storefront after a free translation site failed. A newspaper mistranslation repeatedly misquoted a former president of Kazakhstan as referring to the important issue of “passing gas.” Israeli journalists nearly sparked an international incident when they seemed to insult a Dutch diplomat’s mother in a machine-translated message. Finally, an automatically translated furniture tag contained a racist slur that seriously offended customers in Toronto, Canada.
What differentiates the merely humorous from the cringe-worthy are often the value of the text undergoing translation coupled with readers’ quality expectations. Informal instant-messaging conversations or user-generated content in social media is of relatively low value, so translation errors result in minimal repercussions for even the most horrendous mistranslations. By contrast, legal contracts, financial reports, marketing collateral and application user interfaces usually include text of much higher value that should be translated by human professionals. Text of intermediate value may support a quality level between the two, if the expectations of both the customer and vendor are set appropriately.
When Canadian hockey fans expected quality translation from the French shopping website of their Olympic hockey team, they were sorely disappointed. Machine translation errors irked many visitors, and the team shut down that e-commerce section of the website, foregoing the potential revenue stream.
In contrast, when someone intentionally uses machine translation to simply get the “gist” of a document, and when the alternative to that low-quality translation is no translation at all, they are not nearly so disappointed by the results. When machine translation’s limitations are understood and anticipated, such automatic solutions can be successfully implemented to translate large knowledge bases of user-generated help documentation. Automatic translation can even help facilitate some casual, low-value conversations that would not usually justify an interpreter.
In other cases, legal, financial and political workers are able to comb through enormous volumes of machine translated files — translated behind firewalls using secure systems, not free online tools — to identify key words and select the most pertinent and critical documents, which are then forwarded for higher-quality human translation.
These principles are even understood by Google and other companies that build and market machine translation products. Yes, Google has built an impressive statistical machine translation system, but the search giant involves human professionals to translate higher-value content.
These tips may seem like common sense, but we should not assume everyone “gets it.” As I was so painfully reminded earlier this year, everyone in the content production process must understand the basic capabilities and limitations of machine translation. Unedited, low-cost machine translation can be excellent for translating low-value text and providing the general idea to people who only expect the “gist.” For texts of greater value and for audiences with higher expectations, professional human translation will help companies avoid translation blunders and their costly consequences.