|Thread title||Replies||Last modified|
|Translation technologies - up to date?||0||20:46, 25 December 2022|
|Apertium||6||09:40, 1 October 2015|
|Translation memory||4||18:50, 28 December 2014|
|Wrapping||1||23:51, 15 January 2013|
|Arbitrary change||4||23:44, 26 June 2010|
Last edit: 18:15, 21 October 2013
Hello, the MediaWiki Translate extension includes Apertium among the external helpers for translations. Why is it not available here in translatewiki.net? (fwiw it doesn't seem to be included either in the Wikimedia multilingual wikis. Thank you.
There are no reliable backends available. If you have any, please let us know.
What do you mean by reliable? Apertium seems to be supported by the Translate extension. It is working in my wiki and as far as I remember I didn't do anything to activate it. It just came out of the box. See for instance this combination from English to Spanish. Suggestions from Apertium just show up, and they are helpful.
As I said on IRC, reliable is a service which Translate doesn't keep disabling because of HTTP timeouts or other failures. Even when it was enabled, Apertium was off most of the time on this wiki; IIRC that was the point when it was disabled. If the situation has improved, it probably costs nothing to try and enable it again: do you have any news in this regard?
The backend wasn't reliable enough for translatewiki.net. If there are too many timeouts or errors from the backend, it is automatically disabled. It was disabled too often, so we disabled it before August 2012. We'd be happy to interface with the people behind http://api.apertium.org and hear if they think the service has become more reliable, so that we can enable it again.
Please point them to this topic.
Is there a way to import a translation memory into translatewiki? For example the Korrect 43,000 aligned Welsh and English software translations (under GPL licence).
"The collection was originally assembled in 2004": is that really the best source? Something like that is being worked on right now. The plan is to import l10n from free software projects, as they did, but in much larger scale, so their data should be redundant. Are you available to help importing those projects («KDE, GNOME, KOffice, Mandrake, Red Hat, TuxPaint, Rosegarden etc.»)?
I am sorry, I did this never before. But I could contribute a translation memory for the Ttmserver in an African language (ln) that you cannot find online.
Is it really necessary to wrap like that? Using only half the available space even on a 1024x768 screen means that most of the text is outside view attention area, especially as there's no TOC and the only option is scrolling.
Imho it is better to propose to make a change in the source page to keep all translations consistent.
Change the source page, please if you really think improvements are needed. A staff member will review it before the new version is put up for translation.