User talk:Nike/Archive to 2010-01-01

Time to discuss Maltese Plural

So i wanted to discuss how the plural in maltese works so we can implement some code to aid Maltese translations. As the current system is, the plural function is structured like this:

{{PLURAL$|SINGULAR|PLURAL|}}

which is good in certain cases that doesn't involve definite numbers.

eg. {{PLURAL$|SINGULAR|PLURAL|}} {{PLURAL:$1|Category|Categories}} - English {{PLURAL:$1|Kategorija|Kategoriji}} - Maltese

but when we need to specify a definite number, like for example:

This category has the following $1 subcategories.

in maltese this will have 4 versions

1. Din il-kategorija għanda għandha subkategorija waħda - when we're talking about exactly 1 subcategory.

2. Din il-kategorija għanda għandha $1 subkategoriji - when 2<=$1<=10

3. Din il-kategorija għanda għandha $1-il subkategorija - when 11<=$1<=19

4. Din il-kategorija għanda għandha $1 subkategorija - when 20<=$1<=101

then it precisely recurs as explained in the algorthm you have in the sandbox.

so what we're gonna do? --Giangian15 15:13, 21 July 2008 (UTC)

I don't know. I was hoping one of the alternatives I proposed would be ok :). Sorry for the late answer. – Nike 14:17, 26 July 2008 (UTC)

Actually Gian, 4. should be changed and a 5. added as follows:

4. Din il-kategorija għandha $1 subkategorija* - when 20<=$1<=100 (sottokategorija, to be precise)

5. Din il-kategorija għandha $1 subkategorija u waħda when $1=101 (or 201, 301... ie. recurs every 100)

And only then, it repeats. --Roderick Mallia 10:24, 12 August 2008 (UTC)

Moreover, any code should take into consideration the fact that in Maltese, the numbers change according to the gender of the noun. e.g., sottokategorija (F) = waħda; but artiklu (M) = wieħed. And if the noun is of semitic origin (generally speaking), it itself also changes according to number --Roderick Mallia 10:50, 12 August 2008 (UTC)

Launchpad + redundancy

Hi Nike. I believe you might already know Launchpad, the translation platform. I'll save myself the time to describe the translation suggestion feature they have there (which you possibly already know), by sending you an image :) http://img230.imageshack.us/img230/715/launchpadsn3.png

So, I was wondering, how hard would it be to implement this on Translatewiki? It seems to be a great idea, and would be really useful since many extensions have values that regular mediawiki messages, or other extensions, already have, and suggestions would greatly save time for translators. What do you think? --Waldir 19:22, 27 January 2009 (UTC)

Are you talking manual suggestion by users or automatic suggestions based on other translations (so called translation memories)? – Nike 19:25, 27 January 2009 (UTC)

Automatic suggestions based on other translations :) Manual suggestions would actually double the manual work instead of halve it, if it is what I'm thinking. --Waldir (talk) 00:10, 28 January 2009 (UTC)

It is on the todo list. Implementing it would not be very hard, if someone made a suitable and reasonable fast algorithm we could just feed data on and then query. – Nike 07:36, 28 January 2009 (UTC)

Great! I'll be looking forward to that :) (btw how long is that todo list and which rank is this item at? :D) --Waldir (talk) 10:06, 28 January 2009 (UTC)

See my user page. Siebrand 14:08, 28 January 2009 (UTC)

It is more about the lack of time than length of the todo list. All help is appreciated. – Nike 15:12, 28 January 2009 (UTC)

Thanks

Hi, Niklas. I would like to tell you that the new {{GENDER:}} feature would be a huge improvement for Arabic messages. Thanks for implementing it. --Meno25 15:46, 30 January 2009 (UTC)

another would-be-cool-to-have feature

I just created the google translate bookmarklets in my browser, as suggested at http://translate.google.com/translate_tools to ease the translation of messages. It doesn't get it 100% right, but the accuracy is already acceptable it and saves a lot of typing time. Do you think it would be a good idea to use something like the Language API to prefill the edit field for new translations here in translatewiki, with the automatic Google translation? We would then just have to edit/fix the text, and perhaps there's even a way to send the edited version back to Google so they can improve their machine translation algorithms. Is this feasible? I think it would be great for everyone, especially if we could send data both ways. --Waldir (talk) 14:25, 31 January 2009 (UTC)

Looks like a quick job for someone JavaScript-savy to create such gadget, if there isn't one yet. – Nike 15:25, 31 January 2009 (UTC)

Who'd say? with the kind help of Mike_lifeguard and Splarka, I managed to get something working at User:Waldir/modern.js. It would be much simpler if the message definition had an id, so we could use getElementByID. That is, if instead of something like

<fieldset class="mw-sp-translate-edit-definition">
 <legend>Definição da mensagem (Tasks)</legend>
 <code>Task|Dates|Initial comment|Assignment/Actions/Page</code>
 </fieldset>

we had something like

<fieldset class="mw-sp-translate-edit-definition">
 <legend>Definição da mensagem (Tasks)</legend>
 <code id="msgdef">Task|Dates|Initial comment|Assignment/Actions/Page</code>
 </fieldset>

(of course, "msgdef" came just from the top of my head. any better suggestion is welcome). Do you think it's feasible?

ps - I didn't check whether it is possible to send the edited text back to Google, but it will be my next step. I'll update here. --Waldir (talk) 00:58, 1 February 2009 (UTC)

About the p.s. from my message above: doesn't seem to be a possibility. According to Sparkla it' might be doable with some amount of hacking, but it wouldn't be very pretty. I've requested the feature. --Waldir (talk) 01:50, 1 February 2009 (UTC)

I could add an id, but what about if we want to have multiple message per page someday? There should be pretty good getElementsByClass or something and then loop. Or is this about accessing the actual element inside fieldset? – Nike 09:55, 1 February 2009 (UTC)

You have a point. It that case it is better to stay as it is. It can be easily adapted to support multiple messages per page, but I think doing that right now would create a solution looking for a problem :) I've been testing it and it seems to work fine. What can I do to make it a gadget? --Waldir (talk) 11:44, 1 February 2009 (UTC)

Make a page like Mediawiki:Gadget-googletranslate.js and add it to MediaWiki:Gadgets-definition. – Nike 15:28, 1 February 2009 (UTC)

Thanks, done :) --Waldir (talk) 16:03, 1 February 2009 (UTC)

It doesn't seem to work all the time, because English translation is already pre-filled for MediaWiki messages. Perhaps a button to get a suggestion would be more intuitive? – Nike 16:35, 1 February 2009 (UTC)

Drawing i18ned text in images.

Hi Nike, I have read your blog entry. Commenting it would not work, so I write here. I found it comparatively easy to use vim and sed to replace i18n strings in svg images. Even though having to modify positions on the image canvas due to longer strings, or adapting character height, was tricky. It was all manual work, though some of it quite repitive. Samples are: Commons:File:Autoroute A14 (BE) Logo.svg (and most other signs in the series), Commons:File:Autoroute R8 (BE) Logo.svg (and most other signs in the series), Commons:File:File:A3Swiss.svg, other similar ones, and Commons:File:Griffith experiment ksh.svg on MediaWiki Commons. Maybe marking positions with comments or the like, anticipating what modifications to are to be expected, may make life easier.

May be, there could be a tool that handles such replecements, one way or another, when authors reserve space for strings, and label them appropriately. I don't know enough about the svg format, but believe it would be possible to put placeholders like e.g. $1, $2, …, $N in such images, and have a list of ($i, langcode, string) triples associated, which are used for replacements when an image is served, or precompiled and cashed.

Of course, that is not as easy with other image formats, but many of the types of images that typically have text strings inside, are drawings, maps, etc., which are good candidates for vector graphics anyways. Greetings. --Purodha Blissenbach 07:09, 25 March 2009 (UTC)

The extensions which are drawing images in MediaWiki use bitmap formats. That is a problem. – Nike 07:33, 25 March 2009 (UTC)

Yes. I've seen that from the diagrams e.g. in the language portal pages. I wonder, since making svgs showing the same grid and lines should be rather simple, I believe. Yet, I have no idea about font sets available for inclusion in svg images. Likely, again, so called "exotic" ones are going to make problems - How about asking an svg-expert for help? --Purodha Blissenbach 13:23, 25 March 2009 (UTC)

Translators

Any chance you could contact some language schools and colleges to help practice their language skills by translating from Wikipedia:WikiProject Intertranswiki? I set up this project on english wikipedia (which you are welcome to join) which seems exactly what you also have in mind??? We are drawing up a missing article directory from other languages to translate into english. For this to be a success we need numbers. Publicizing that we need translators in any language I think is an important step to make. Have you browsed French, Spanish and Italian and German wikis for instance. There are several million articles which we could benefit from having them translated into english... Please could you organize a way to contact various language schools and classes to attract them to this project.