In classic sci-fi movies, contests between machines and humans almost always end with the machine's demise due to the stress of competing with superior human reasoning power. These days, machines beat humans on game shows, computers win at chess, and the quality of machine translation (MT) improves every year.
Let's take a look at three types of translation techniques, their definitions, and what content they are most appropriate for:
- Human translation
- Pure machine translation
- Machine translation with human editing
1. Human Translation
A professional linguist (most often, an in-country native speaker) reviews your project and, using guidelines agreed on beforehand, translates it to the language you require. The goal is to speak to your audience in the most natural, effective way. You can expect human translations to be free of idiomatic errors and to flow naturally and fluently.
Advertising and marketing projects can be "transcreated," which means using your headlines, copy, scripts, and product names as the starting point. Your material is then creatively translated into culturally sensitive language that will appeal and make sense to your global audience.
Best candidates: Projects that need to convince, persuade, build trust, inspire, educate, entertain, or sell your product. For example:
- Print and broadcast advertising
- Marketing and branding materials
- Store signage
- Social media
- Product and brand names
- Website content
- Multimedia (e.g., Flash, voiceovers, etc.)
2. Pure Machine Translation
Pure machine translation is a computer-generated attempt to reproduce the language reasoning that human brains perform. Because translation is all about interpretation of meaning, our brains perform loads of cultural assessment, analyzing nuances and expression to fully comprehend language. Our brains are able to assess these nuances and translate them. Machines have not yet been able to do that, even with the simplest types of text.
Rules-based machine translation dates back to the Cold War, but today there's a new way of approaching machine translation called "statistic-based translation."
To perform statistic-based translation, a search engine delves into the billions of words and word pairings on the Web and produces statistically good matches for the way things have been said in one language with the way they've been said in another. And tools are available that apply some grammar rules to the translated material, producing some fairly decent results.
However, you'll still find thousands of often-hilarious examples of the pitfalls and limitations of machine translation on restaurant menus and store signs—and sometimes, in important business communications—around the world.
Best candidate: Personal use. Here are some examples:
- Looking up a word or phrase and translating it into your native language, or vice versa
- Travel aid—translating restaurant menus, directional signage, maps, and more
- Just for fun—tweets, Facebook updates, quick notes to friends or family
- Getting the general sense of a short piece of text when exact details are not important
3. Machine Translation With Human Pre- and Post-Editing
This hybrid is akin to a cyborg: It's a more serious, controlled machine-translation software used by professionals with lots of up-front prep work.
Here's how it works. A linguist goes through the project first, then "trains" the machine-translation engine to translate properly. For example, the linguist will feed long lists of words with double meanings into the software, essentially tweaking the software's rules to tailor the localization to a specific project or client.
After the material is sent through the software, the linguist will look through the first few thousand words to check for mistakes and, if necessary, will retrain the software to interpret rules correctly. The material will go through again and will be reviewed by post-translation editors who make minimal changes to ensure that the material is technically accurate and understandable to readers.
Those translations will definitely not be at the level of fluency of human translation. But if you don't have the budget to localize those 10,000 data sheets, human-aided machine translation could be a great solution for you.
Best candidates: Large-volume projects of more than 500,000 words
- Projects that require a very large volume of words to be translated and therefore justify the considerable setup time
- Straightforward text
- Technical manuals and data sheets
- Safety instructions or documents that must be posted by law (or anything that needs to be accurate, but where style isn't the first priority)
- Customer reviews on your website
- Large internal documents that are not consumer-facing
* * *
When making decisions about which localization method to go with, give careful consideration to the type, size, and audience of your project; and, of course, keep your customers in mind, too.
One thing's for sure: There isn't a machine on earth that can help you make those very human decisions.
You may like these other MarketingProfs articles related to Content:
- Build B2B Marketing Trust With Evidence-Based Content: Melanie Deziel on Marketing Smarts [Podcast]
- Webinar Invitations: Examples and Best-Practices
- The Cost of Poor Business Writing
- 12 Reasons User-Generated Content Is Important for Brands [Infographic]
- Why You Need a Branded Podcast (And How to Create and Brand Yours)
- Five Trends Fueling the Rise of Visual, Data-Driven Storytelling [Infographic]