Loneliness of the long-distance linguist. : localization

Found in translation.

KieranS — Thu, 07 Sep 2006 09:46:00 GMT

Though languages can be interestingly different from one another, the majority of work in modern linguistics has been dedicated to showing that the differences are only cosmetic; at heart, all languages are expressively equivalent. I know this is true, because every time someone from [big name theoretical linguistics department] comes up with a new model to describe this, a whole bunch of people who used to be at [big name theoretical linguistics department] run out and apply it to their favorite languages and show how it works. :)

But in software, things work differently. The kinds of linguistic differences that are meaningful to linguists are not the same as the kinds of differences that are meaningful to software developers. And for a linguist working in the software industry, that's where things really start to get interesting.

All languages are expressively equivalent. What do linguists actually mean by that? They mean that although some things can be difficult to translate, there is no semantic content that can be expressed in one language but not in another. You may have to say it a different way, using different vocabulary, syntax, gesture, intonation, and so on, but if it can be said in some language, then you can say it in yours. From the point of view of a theoretical linguist, this is a pretty deep contention, because it shows something about how the brain works regardless of the language that someone happens to speak.

Well. Full disclosure: From the point of view of some linguistic anthropologists, the above contention sounds pretty wacky. But let's pretend that theoretical linguists are right and the body of work supporting this from the last half-century is really on the mark. That's the kind of background I come from and I find the work pretty convincing. So what's the problem? The problem is that in software, it doesn't actually matter if it's true. It's true, but it's not usefully true. Because in software, you have to design stuff, and when you start designing stuff you run into all kinds of practical constraints that theoretical linguists don't need to care about.

Let's look at localization as one example. On the one hand, it works out pretty well that all languages can figure some way to express opening, saving, printing, and closing files (even if sometimes there's some discussion about the best way to say it).But how far does that take us? If you're developing an application, then you still need to figure out how to design your UI so that all relevant localized strings fit within it. It's great that text can be localized into several expressively equivalent languages, but if you're an application developer, you're not interested in the extent to which languages are the same in this regard. You care about the range of differences between them. Because your application needs to plan around exactly that set of differences.

You can find analogous examples all over the software space. It is crucial to grasp the range of sounds across languages in order to do good work in speech. You need to understand the ways in which different languages can segment written text into morphemes, words, sentences, and discourse segments before you can really plan great content rendering support. In all of these cases, really good multilingual work starts by documenting the range of variation as closely as possible; it is only by delineating the differences that the necessary design generalizations start to emerge.

Lexical diffusion, software style.

KieranS — Thu, 17 Aug 2006 10:53:00 GMT

The Microsoft Community Glossary - Linguistic Partnership Project allows members of a speech community to work together to create and standardize technical vocabulary for their language. Community Glossaries can be an important step on the road to software localization for customers in regions that are newer to technology. Thus far there are online Community Glossary projects for languages all around the world, including Greenlandic, Kannada, Māori, Thai, Vietnamese, and several others. Interested participants work with a local moderator to decide on the appropriate terminology to be used in the software space.

So does it work? That's an interesting question.

New vocabulary is introduced into a language in several ways. Sometimes, rarely, someone well-known or otherwise influential coins a word or phrase that happens to stick. This can happen on many scales, ranging from the popular kid in a group of friends all the way up to celebrities who get media attention. Sometimes a new technology is introduced and some brand name associated with it becomes the generic term, as with kleenex, xerox, and more recently and much to their apparent chagrin, google.

More often, though, new vocabulary gets introduced when speakers of one language come into contact with speakers of another. English has borrowed vocabulary from virtually every language with which it has come into contact. No matter how government language authorities try to impose standards around vocabulary in advertising, media, and school curricula, such efforts have not generally been successful in achieving results in the real speech of real people. Mexican kids still speak Spanish during Little League games and the French still end the work week with le weekend. It's just how language works.

That's not to say that all standards efforts are ineffective. When a particular term catches on and its usage is popularized, it becomes a de facto standard for the speech community that uses it, and its very popularity both propels it forward and also catalyzes its eventual decline. Vocabulary choice is one way that speakers define themselves either as part of or against the speech communities they inhabit.

What is true for colloquial vocabulary is also true for terminology in the technology space. Terminology persists when people use it. Which brings us back to the Community Glossary efforts.

If efforts to identify and standardize technical vocabulary are driven solely by official language authorities, I don't think they will be successful in the long run. A word is only as present as the community that uses it. It is only by involving the target user community that standardization efforts can hope to identify relevant and enduring terminology, and the most successful products will be the ones that create a user experience with that terminology in appropriate ways. I'm very excited about the Community Glossary project because it lets real speakers identify real vocabulary that actually works for them, and it creates a lexical resource that can be valuable in language preservation efforts for endangered languages.

Anyone who believes that a Community Glossary effort would be beneficial for their language can work with their local Microsoft subsidiary office to request one and get the process started.

Lights, camera, action!

KieranS — Fri, 04 Aug 2006 01:44:00 GMT

For everything you ever wanted to know about Windows globalization and localization, or at least for one or two things, you might like to check out this Channel 9 interview. It features Michael Kaplan, whose blog you may know; Jeff Allen, our director of localization services; Alp Turkmen, who is driving our business direction in the Windows international space; Claus Juhl, one of our localization engineers; and me. Michael and I speak about globalization issues, with a heavy focus on the Locale Builder and associated ways that we'd like to open up our infrastructure to accomodate all kinds of community solutions, and we're about 45 minutes in. But don't skip! There's great information throughout.

If you watch, please let us know what you think! And yes -- there are closer to 7,000 languages in the world than 2,000, for those of you playing along at home.