Wall Street Journal Watches Website Language Management Industry; Illuminates Opportunities, Common Pitfalls

January 9, 2008

During the past three months, The Wall Street Journal has turned a keen eye to the topic of translation, and has paid increasing attention to Website Language Management (WLM). The newspaper has educated its readers on the shortcomings of available technologies to deliver high-quality translations.

The stories, summarized below, illustrate noteworthy developments that every global company -- or company hungry to connect with international markets -- should know about. Unexpected challenges may face companies that are trying to reach new global markets, these stories say. They also showcase why only a provider using skilled human linguists to translate content can deliver the high-quality, in-language experience global audiences demand.

On Oct. 22 The Wall Street Journal published a story in which reporter Peter Loftus explored global corporations’ use of software-based translation -- also called “machine translation” -- services. Machine translation is making it easier for companies to do business in other languages, he wrote, but “(h)uman input is still needed to ensure complete accuracy, especially for material that customers will see, like company Web sites or product packaging.”

It’s well-known that automated machine translation is cheaper than translations provided by trained human linguists. Ford Motor Co. uses such software to convert English manufacturing instructions into Spanish, German and other languages. These documents are used in-house only, and do not publicly represent the Ford brand. This is likely because of the universal concession that machine translation is not -- and will likely never be -- 100 percent accurate.

According to the story, Ford line workers in Spain, Germany and other countries correct these machine-translated documents as they go, ensuring accurate manufacturing. These workers are experienced in both vehicle assembly and deciphering machine-translated content, so this solution can work in this unique situation.

A marketing officer for one such machine translation provider, was quoted in the Oct. 22 story as saying: “No one thinks you can depend solely upon a machine. I do not believe you can ever take the human out of language when it comes to publishable (content).” In fact, every translation industry expert quoted in the piece agreed: Humans are needed to translate or polish content intended for public consumption.

In the Dec. 21 edition of The Wall Street Journal, technology columnist Sarmad Ali further explored the limits of machine translation. As the demand for international communication increases, so does the demand for free, user-friendly online machine translation services, he wrote.

To satisfy his curiosity, Ali “tinkered” with four such machine translation services. He processed three breeds of text through the software: content with a conversational tone, news stories and legal documents. While he found these services easy to use, Ali discovered that “all of the services did a terrible job with metaphors and other figurative uses of the language.” The services “were better at translating everyday phrases, but even these sometimes came out missing a word, or were scrambled.”

Ali, a fluent Arabic speaker, took particular interest in the four services’ Arabic translations. These were “rife with syntactic and semantic errors -- from the merely too-literal to the laughably bad” and “would have been nearly impossible to understand were I not fluent in both (English and Arabic).”

Improvements in machine translation will be slow in coming, Ali wrote. “Use with caution,” he warned in the last sentence of his column.

To further illustrate the limits – and potential risks – of using machine translation, Wall Street Journal blogger Ben Worthen documented a November incident in which an Israeli journalist used software translation in his correspondence to the Dutch foreign ministry. The end result? Outrage from the ministry, and embarrassment (and a lost interview) for the journalist.

This journalist, who does not speak English, was asked to provide a copy of the questions he would ask the minister on his upcoming trip to the Netherlands. Rather than ask an English-speaking colleague to translate his questions, the journalist processed his document through a popular online machine translation service. The software confused one Hebrew word with another, and the question was mistranslated into a request to know where the Dutch minister’s mother slept.

“How could this e-mail possibly have been sent?” said an official with Israel’s foreign ministry, quoted by the Jerusalem Post, as cited in the WSJ blog. “Sure he can’t understand many of the questions, because the English is so bad. But he is being asked about the sleeping arrangements of his mother!”

A website translation solution such as MotionPoint’s solves these problems and anticipates innovations. Not surprisingly, many of these issues arise from expensive legacy methods of translation ... which spawned a host of machine translation providers (and deliver inconsistent results).

The translation industry grew up around document translation, whose process is 90 percent translation and 10 percent translation workflow technology. Translating Web content is a different animal, in which the translated text and images are often deeply buried in all types of different technologies. The key to this lock is 75 percent technical – meaning, the website translation faces seven times the technology hurdles. With a WLM-powered site, humans perform all translating of text and graphics, but they are empowered by a team-based, Web interface packed with thorough proofreading, glossary and QA features.

Most companies that specialize only in traditional document translation don't have the technical know-how (or Web-based experience) to meet such demands. And as demonstrated in these Wall Street Journal articles, machine translation falls short in quality.

For companies looking to communicate with new international or multi-lingual markets, the technological and language-specific challenges are many. To sidestep these worries and deliver a high quality, human-translated multi-lingual site, partnering with the right WLM provider is key.