Ryan Coleman, another friend of mine from the TorCamp community, led a discussion on language translation and the impact on the sort of interacting with the global community due to the premise of wikinomics. Although it’s easy (and arrogant) for those of us who are native English speakers to just ignore other languages and pretend that everyone speaks English, the fact is that if you’re message isn’t well-understood, you’ll end up losing business or creating inefficiencies within your organization. At this point, translation services is a $10B business worldwide, and growing.
He gave some examples of evaluating the context and content to determine whether it needs to be translated, and the degree of care that needs to be taken, before going through the different options for translating your business materials.
One option is to crowdsource your documentation: have your user community write the manual for you. This requires a passionate user base, and can be unpredictable in terms of timing and coverage, as well as of inconsistent quality.
Another option is machine translation, but as you’ll know if you’ve ever used Google Translation, the quality can be total crap with low-end solutions. There are high-quality (and higher-priced) professional systems, but these require extensive training and still require review of the output.
Another option is to use internal resources, namely your own staff, who presumably understand your products and services, but who are now diverted from their usual job which tends to create a high cost of lost opportunity. Since these are not professional translators, the quality can also be questionable.
Professional translators are the final option, and best for high-quality, consistent translation. They can use tools to store translated phrases so that there’s a translation memory of a document; when a document changes, only the changed portions required re-translation. The downside, of course, is that this is very expensive, and the initial translations can be time-consuming especially if you have a lot of specialized terminology that the translator needs to learn.
There are a number of hybrid approaches that combine these options; all of them will combine people, process and technology in some proportion, and the ultimate choice will depend on both the content and the context.
Ryan listed a number of other points to consider:
- Synchronization between versions, including maintaining dependency relationships
- Location and access to content repository
- Workflow and time sensitivity of translation, including proofing/review cycle
He had some thoughts on what’s happening between translation systems and content management systems, particularly for large websites that must be maintained in multiple languages. In the past (and likely still a lot currently), a content management system would just spit out a document to be translated, then accept it back in afterwards, without any real sense of how the translated content should be handled. Wikis, of course, are even worse since it’s less mature responsibility and there’s not, in most wikis platforms, any considerations for maintain multi-language versions of a wiki.
Ryan’s company, Clay Tablet, has created a piece of middleware that sits between the different types of translation systems and the content management systems, whether the translation is being done by a machine translation system or a company that provides human translation services.
That’s the end of the formal sessions of Enterprise 2.0 Camp; it’s 2pm and we’re decamping, so to speak, to the bar across the road for lunch and a continuation of the conversations.
Thanks for reminding me that an active blog is better than…oh, well, mine. A very faithful account of the conference, including sessions I missed.
As a lark to reinforce the translation argument, I cut and paste your “total crap” Google Translator paragraph into Google Translator – starting from English to French, and then back. An ideal machine translator of course would render perfect results.
You get this:
“Another option is machine translation, but because you will know if you ever employed the translation of Google, quality can be total shit with solutions of low-end. There are the professional systems of high quality (and more expensive), but those requires the wide formation and always requires the examination of the output.”
Surprisingly good, although the devil is in the details. The phrase “quality can be total shit with solutions of low-end” is the kicker. Professional translators can parse through idiosyncracies and slang and render such phrases properly, which is why Ryan noted, they’re worth the big bucks, even if only for phrases such as this.
Translation requires both syntactic and cultural fluency. As a result, machine translation is as fluent as a kid with a large dictionary and only the faintest of clues of what they’re doing with it. You can try to make heads or tails of what pops out, but since it’s not based in contextual understanding of language, the essential flavour of the original is twice-removed from the source, and thus the results are hackneyed or worse.
A perfect example – the paragraph above – deliberately colloquial in terminology to make a point – translates Eng/Fr and back to this:
”
The translation requires the syntactic and cultural control. Consequently, machine translation is as flowing as a child with the large dictionary and only weakest of the indices of what they do with him. You can try to make heads or tails of what noises outside, but since it is not based in the contextual arrangement of the language, the essential savour of the original two time-is removed source, and the results are thus folded back or worse.”
Understandable, kind of, but a real effort.
What is interesting (and notable about your blog postings, actually) is that straightforward language can easily be parsed to reasonable degrees of accuracy, whereas flowery nonsense (more my style, alas) is mangled beyond belief.
Too often, people privilege the flowery as intellectual, and to bad effect, IMO. It’s harder to understand, even without machine translation. It’s too high-context, high-touch – great poetry if done right, perhaps, but as a means of delivering a simple point, very inefficient.
So, it’s nice to see a very appropriate and factual account of today’s happenings. It’s like I was there. And I was, but your liveblogging’s actually beneficial in supplementing my already extant understanding, which makes it particularly valuable. Thanks!
Michael, thanks for your insights on translation. I’m puzzled over how “extensive training” becomes “wide formation”; my French isn’t good enough to understand why that particular translation occurred!
The writing in my live blogging from conferences is usually pretty plain language since it’s mostly stream of consciousness (and my consciousness isn’t all that flowery). Other writing that I do for clients sometimes isn’t as plain, and I take your point that maybe it should be; in fact, I made the point in an earlier post today that someone’s website was somewhat incomprehensible because of the overly fanciful language that they used.