Featured Video
This Week in Quality Digest Live
Innovation Features
Trevor Blumenau
Inexpensive wireless pick-to-light systems put warehouse productivity in reach for everyone
Tab Wilkins
Resolve to check out these resources
Dirk Dusharme @ Quality Digest
Manufacturing, urgent urgency, and a robot took my job
Laurel Thoennes @ Quality Digest
Scratch taxing the rich to feed the poor
Steven Brand
From robotics to AR, new technologies push the envelope

More Features

Innovation News
125 strategies to achieve maximum confidence, clarity, certainty, and creativity
MIT awards more than $1 million to organizations creating greater economic opportunity for workers
Berkeley Lab and Magic Leap Inc. scientists create widely controllable ultra thin optical components
$79 device delivers dedicated neural network processing to a range of host devices
Drip irrigation targets the plant and not the soil
New approach uses light instead of electricity
If you want to understand a system, try and change it
Components will be designed from the onset to inhabit multiple configurations during service
More than seven billion lives may depend on it

More News

Graham Ross

Innovation

Machine Translation’s Effect on Global Business Communication

Humans will still be involved, but the service will be cheaper

Published: Monday, October 2, 2017 - 11:01

Why doesn’t machine translation (MT) work for all languages? There are currently more than 7,000 languages worldwide. If each of those had four dialects, that would be more than 28,000 language variances. To offer a perspective, Google can translate 37 languages. If you take into account integrating artificial intelligence (AI) into each of the 7,000+ languages for MT, we are a long way from being able to translate 100 percent by machine, if ever.

Before beginning my career in the language service industry, I spent 25 years in global sales and marketing with manufacturers of maintenance, repair, and operations (MRO) equipment. I believed at that time that a Spanish translation meant replacing one English word with the Spanish word of the same meaning; translating was black and white.

However, once I started working in the language service industry, I realized how wrong I was. International languages have different grammar structures and a wide variety of dialects within a language. For example, French has 28 dialects while Spanish has a mere 11. A common issue arises when the translator lives and works in a different area of a country than the client’s reviewer, each with a distinct dialect. You have two interpretations of the same translated document, and both are right. The translator translated it in her dialect while the reviewer is reviewing from his prospective client’s dialect. The translation is good but different words and phrases were used to say the same thing. At this point the client must choose which dialect he will use to communicate the company’s story in that market.

I’m an old manufacturing guy in a very exciting industry at a very exciting time. The internet is on fire with news about AI being integrated with MT to produce translations that will be either free or of minimal cost. Is that true? Yes. And no. It will depend on the nature of the material and the client’s goal within a particular international market.

Machine translation has come a long way since its inception and continues to improve with each passing day. Still in its infancy, AI eventually will be integrated into MT with very good, but limited, results. MT will work well where the text is factual or repetitive. For example, MT will do a good job with parts lists, assembly instructions, troubleshooting guides, data sheets, and forms. Although MT will do this well, it still will need to be reviewed by a human translator to ensure it is correct and coherent.

Where machine translation doesn’t do as well is with text that conveys a descriptive message, one that conveys the importance of features and benefits, tells the company story or philosophy, or presents leadership biographies. In short, anything with a human thought process.

Google, Microsoft, and Amazon would have you believe that all you need to do feed the text into the machine and voilà! Instantly out comes the ready-to-use translated text. That would be nice but it is unrealistic.

The fact is that MT will be one of several tools that will be used by language service providers (LSPs) to provide their clients with professional, accurate, and appropriate translations for a targeted market. The fastest growing language process is a byproduct of MT called “proof and edit machine translation (PEMT).” A human will proof the MT text to ensure accuracy and edit/fix the translation as necessary.

For example, a company sends a manual that needs translating into into French to an LSP. The manual has part numbers and lists, model numbers with descriptions, product use information, and the value of the features and benefits defined. The LSP will analyze the manual and provide a quote for the service. The quote will itemize an augmented translation process where parts of the manual are machine translated, and other parts are translated by humans. Whatever the percentage of machine translation to human translation, all of it will require editing and proofing to ensure a high-quality translation that will professionally represent and protect the brand in the target market(s).

One of the major benefits of MT is that it will increase the translator’s productivity, which in turn will decrease overall costs. It won’t be free, but it will be less than you are paying now.

Discuss

About The Author

Graham Ross’s picture

Graham Ross

Graham Ross is the business development manager for Transatlantic Translations Group.