4 ways of not doing multi-lingual twitter
Us europeans have a hard time. When we twitter amongst ourselves, between and during our lunches, our hard-working English speaking friends find it hard to understand what we're talking about. So I looked at 4 ways of changing how we twitter to bridge that gap.
1. Switching to English
Most of us speak English pretty well, and switching to twittering in English might be the easiest way. Some of our European friends may not like that though, and I've experienced severe peer pressure when switching to English for the briefest of moments during SXSW.
2. Adding a language notifier in tweets
I've built language detection into my diy twitter client at http://twitstat.com/m/. It works with the Google AJAX Language API, and it actually does a pretty good job at detecting language. So you'll get tweets that say '[nl] was even aan het lunchen' or '[en] that was a great lunch'. Some of my followers didn't appreciate the ugly brackets though, and as long as there is no standard way to filter on these tags, it's not very useful. An American reader will still be confronted with the [nl] tags. Unless of course he uses http://twitstat.com/m/, in which case he can hide all tweets that contain [nl] or [fr]. Or even the word lunch, to get rid of most European tweets altogether. ;-)
3. Tweeting English in a separate account
Another option is to use one account for your native language, and one english twitter account for your foreign relations. This turns out to be very unpractical and will lead to many mistakes. I've added the option to automatically post all my English tweets to my foreign account, but you'll get funny stuff happening like replying in English to a tweet to your native account, which will then have the wrong sender in the @replies of the receiver. Confusion.
4. Automatic translation to two accounts
The final option I tried was to see if automatic translation via the Google Translation API could be viable. A dutch tweet entered in my favourite twitter client was automatically translated to english, and posted to my english account. Unfortunately, this has several unwanted effects. A reply will be translated as well, so the recipient will get 2 replies: One in dutch from my regular account, and one in English from my translated account. The other drawback is that the Google Translation API actually is complete rubbish and will only translate the most literal of sentence correctly. It really is quite poor. Check out my auto-translated feed if you don't believe it.
Conclusion
The only good option I see is to have the twitter client do the language detection, and then either putting the [en] tags in the tweets, or in another field that can be accessed via the API. If all twitter clients would be adapted to this, anyone could choose to filter out the languages that he doesn't want to see. This only works if all Twitter clients will add this language tag, and offer language filtering.
Comments
Interesting list of points. Now if you will allow me to *again* make my case regarding multilingual twittering ;-) (Michiel and I have been having this discussion for quite a while now on Twitter)
The way I see it, the desire for multilingual twittering is but one of the things that signifies that it's time for the next phase of the Twitter platform as a whole. We need to start separating the data and the metadata; the tweets and the metatweets.
If we could simply add a tag (or language notifier) to a tweet specifying it as being in a certain language, not visible to readers but machine readable by a twitter client, then the problem is solved: readers could easily specify in their Twitter clients whether they want them to display all your tweets or just the ones in their native language.
'Easily' of course meaning: in the case that all Twitter clients support such a method of interpreting metadata like that. This is big step, but if you think about it, it's also a very logical one for the Artform/platform itself to take.
I see it as very analogous to the MIDI protocol in digital music: a MIDI message is usually comprised of a status byte and a data byte: one signifying what kind of message it is (note on, pitchbender, Delay Lama vowel etc.) and the other actually containing the data. The same would be very logical to do with tweets and metatweets, or whatever you'd like to call them.
Apart from language, you could also put hashtags like #leweb in the metatweet, allowing for private twitter streams that don't swamp the timelines of those not interested in some guys not talking about applying 2.0 phenomena to money systems themselves. But I digress ;-)
Cheers, Steven aka @kruithoph
Posted by: Steve Crosswick | December 17, 2008 08:28 PM
Steven,
Thank you for your comment. I must say I agree with you. :-) Option 2 is definitely the best way to go. If only the API would build this in, then twitter clients could start inserting the right data and filtering the output.
<lang>nl</lang>
<hash>leweb, le web</hash>
is all it takes..
Michiel
Posted by: michielb | December 17, 2008 08:34 PM
O darn, us Dutch always so convivially agreeing...
I'm sure something like this will get built into the API, but that doesn't keep us from trying to hack it beforehand ourselves, of course. You already have a twitter client, so lemme see, how about...
1. provide the user with a standard tweet entry field, and a simple radio button for Nederlands or English.
2. encode the radio button output into a second tweet, and post it to a second account (eg @kruithoph_meta)
3. devise a method of reliably interpreting the two tweets as being affiliated, perhaps by simply using a timing window;
4. decode the metatweet and filter according to user preference.
Hmm? :-)
Posted by: Steve Crosswick | December 17, 2008 09:15 PM
Hmm.. That could be done although it may be easier to store these things in the twitstat database, that makes it faster, easier to look up, and costs less API calls.
Hmm.. ;-)
Posted by: michielb | December 18, 2008 07:50 AM