Replacing Identical/X templates
Can someone from the translators shed some light how these templates are used (in the sense of your workflow)? How could they be replaced and avoid the manual work?
I think they serve two purposes:
1) Same messages could be translated at once. 2) They could be used to check translation consistency.
Bots is definitely suited to maintain Identical templates. Message variations are regularly minor (capitalized/not capitalized first letter), single/plural, format variation (colon. ellipsis, parameters at the end), etc. May be special extension or Translation Memory improvements could do job too.
The 'suggestions from translation memory' are sometimes enough. But I do check the Identical translations quite often, especially when an English word can be translated with two or more different words (or grammatical variations of a word) in different contexts. In these cases I find it useful to have an overview of all variations on past translations, not just two. Even if there are not more than two translations listed in the Identical templates listing, the layout shows everything together, which is quick to understand. At the same time I do sometimes adjust some of the other translations for consistency. It is occasionally (not often) useful to be able to look at other similar messages which haven't yet been translated, to gain a sense of how this message is used in different places.
I used to use identical messages to translate similar messages at once. However, nowadays I try to do this for core messages only, because I don't want to start translating extensions and other projects when it could be many months before I have a chance to complete them (as recommended by translatewiki.net).
I would also find it very useful to have a quick way to list or read all messages using a particular word. At present I load all translated core messages and all translated extension messages in separate tabs (and sometimes the untranslated ones too) and then use my browser's 'find' tool to read all the possible contexts of a word, and all their translations.
Gathering from these comments, would it be enough if we had good search that would allow
- searching only from selected namespace (already supported)
- searching only from selected language
?
Another thing that is possible to do is "Search for similar messages", implemented using the translation memory. It could complement the search functionality above.
Comments?
I'm all for "similar messages". Should of course be based on the English source message, not on existing translations.
I like your suggestions. For your first option of using a good search, would the results be displayed on a normal 'search results' page, with the page title and some of the page text printed? This would be useful in searching for a particular word or phrase, if we could search in the English language messages only, or in English plus xxx language only. Or I could search for the English word (or phrase) or translated word in the xxx language messages only, on a separate tab. Or were you thinking of having a layout for search results that was like the translation tool layout? That would be brilliant. Could it search in both of all the English source messages and all the xxx language messages at the same time? That would save a lot of time doing searches using my browser 'find' tool.
'Search for similar messages' should also be useful. Even if it doesn't quite do all that the current manual system could do, it's a price worth paying to eliminate the time to create and maintain the manual templates.
I forgot to mention in my previous post that one of the uses for searching for a particular word is to check that a particular xxx language term isn't already in use to translate a particular English term, so as not to have one xxx language term being used for two different English terms.
I don't think that search will help in cases where commons word will be used (name, user, edit, etc).
As about translation memory letters variation should be very strict (only punctuation marks, -s, -es for plural). And translation memory is not updated properly updated in all cases.
It'll not also help to refer to set of similar messages (Update, Updated; Load, Loading; Yes, No, etc).