We're running a multilingual experiment on this wiki. Some pages are multilingual.

Not everybody appreciates multilingual pages, however, and currently there is no easy mechanism to hide text in other languages. The only method available involves CSS and setting of some parameters. This means that toggling languages on and off is not simple, and it means that even if you hide other languages, the text is still transmitted, slowing down the loading of pages.

When to add another language

This is why I suggest the following criteria for adding text in other languages to a page:

  1. If the language is already multilingual, just do it.
  2. If the page is one of the SocialPages, do it.
  3. If the page is one of the main pages such as the ones linked at the very top of the page (SiteMap, RecentChanges, RecentNearChanges?, HowTo, Blog), then don't do it.
  4. If the page is a category page, then don't do it.
  5. If the page belongs to a category that contains mostly monolingual pages, then use only the predominant language.

These rules are of course just an approximation of what I believe would be a good separation between multilingual and monolingual pages.

Use the MultilingualExperiment cluster

If the page is multilingual, please put "MultilingualExperiment" on the first line by itself. This will put the page in the MultilingualExperiment cluster. Visit the page to see RecentChanges for the pages in this cluster only.


I haven't been aware of the technical aspect, the difficulties it makes to filter out languages and the slowing down effect. It should be a usefull MultilingualExperiment here and not cause dammages. So technical reasons would for the moment prevent the application of multilinguality on base-information pages like HowTo, therefor you split them. Again I understand, and better than the last time (when I had started multilinguality on NoLeader). This concerns the reading of the page. At present a wide application of multilingual pages for technical reasons might slow down the wiki and cause dammage. I see a difference on editing such pages though. When I read the same content in different languages which I do understand there is more information, more precise information about it. A multilingual text is a refactoring of the monolingual origin. It will even more when we learn better to BeBold when translating. There is still a hell of a lot of codework coming up as the sychronisation of translation requires a mark up system of changes made in one translation but not in the others. With patience, step by step. Language filtering, also in the edit mode, mark up of translation changes, not now yet, maybe in closer future. As far as the experiment is concerned: Certain multilingual people would rather edit a content on the multilingual version page than on any on the monolingual version pages and they would do it with better quality there. I would therefore like to continue running certain pages of different types as well mono as multilingual. To feel the difference. I started with NoLeader and NoLeaderMultilingual and I would like to try on HowTo and HowToMultilingual as well. If heaven doesn't fall on my head I'll care for synchronisation between the both and I promise that it won't get too many. Just as an experiment, to give everybody a feel how it could look like. And to check out if multilinguality on one page augments the intelligence input somehow. Would it be ok for you if I recoverd the multilingual version of HowTo (version 7, en, de, fr) and put it in the MultilingualExperiment cluster as HowToMultilingual?

I like this approach. We have some monolingual pages for the casual visitors, and we continue the multilingual experiment nonetheless. Thanks!

Personally we are higher organisms, mamals to be precise. Mamals live and survive because they are capeable to react on changes in their sourounding in an adequate time. This (simplified) involves three time-consuming proceses. First your sensors (senses) transmit the information about the change to your brain (or to its extention, the spinal cord). In the brain the information is processed and a "decision" to react is taken, and then (for examlpe) a signal to the relevant mussles to contract or release is sent out. When it arrives and is executed you react. The speed of a reaction is determined by

  • the distance from the sensor to the brain / the distance from the brain to the mussle
  • the speed in which the information travels through the nerves
  • the time it takes the brain to take a decision.

A collaborative real-life group may (rather temporarily) act like an "organism". A symphony orchestra, a squad of firefighters. Also here the determining qualities are physical distance and how fast the information travels though its medium. Nerves are compareably slow. Acoustic information - medium air (mostly) - travels with mach 1. Visual information travels with the speed of light.

The hive-mind is a permanently collaborative online-organism of global size. Mainly the information in it travels through cabels with the speed of light. The hive mind is multilingual. The most decicive aspect for information to travel through it is the time it take to translate the information into its different languages.

Imagine you burn your finger on a hot iron, but your finger speaks only English. Your brain inspite speaks only French and for the information to pass from the sensor to the brain in order to react and draw back the finger it would take you to wait for ChristopheDucamp to translate it. Pretty lame organism, I'd say, not likely to survive very long, eh?

About precision: A known funny game is to take a text, let is automatically be translated to language A, the result to language B and the result of that back to the original language. Its contents will be messed up. That is because natural languages do not follow logical rules and computers can't really deal with unlogical stuff. They can deal nicely with a logical language as lojban.

A precise automatic computer translation from a natural language into another natural language will not be possible. A precise automatic computer translation from lojban into any natural language is possible. So we have to translate a good idea to lojban and then we can automatically distribute it over the entire hive-mind.

Oh sure! It will not work when putting the translation of all languages on one page. All we need is are pages that seem to have them all on it. How to solve that technically? Somehow using anchors, no? I mean, Alex, would you be interested to go on in that direction? With not all the load on one page. I wasn't aware of the technical consequences, I started the experiment the easiest way that came to my mind.

Assume that super-translation-lojban exists, assume making multilingual pages is possible.
An idea comes up worth the afford ;) It's written in english. (Now comes the problem:) It's translated to super-translation-lojban - slt. Once that is done a translation robot translates the text to multiple languages.
xtof likes to translate. His default settings are French and English. When he watches the page he sees the English original, he sees the information that the text has been translated to slt. He does not see the text in slt, as he doesn't speak it and therefore he hasn't activated it. He could though. xtof also sees the automatic (pre)translation of the text to French. It is marked as such: This is an automatic translation never touched by a human. xtof reads the pretranslation and understands the text, it is simplified, it has this stl-taste but it transports the contents far more precise than an automatic translation directly from the original natural language would.
xtof edits the text. Nice edit mode. Sentence by sentence. xtof makes the translation good French. He's ready with it in a minute. Once saved the text is marked as translated by a human.

To do

look for an artificial language that being automatically translated into whatever natural language gives the most precise results.

Partial translation relates.

Define external redirect: RecentNearChanges