When I tapped a section under offline state, it failed and a caution like above popped out (I don't use automatic translation at all as I've said in Prenvented publishing because of "unmodified machine-translated text", but it's not the matter this time.) The problem is, even after I came online, the section which said "Automatic translation failed" can't be edited anymore. Just saying the same caution. Is it possible to complete the translation in such situations?
Jump to navigation Jump to search
Reply to ""Automatic translation failed!""
Reply to "this tool is apparently broken"
Reply to "File name prefix: pls unlimit adding "Wikipedia:" on jawiki"
Reply to "Reference problem"
Reply to "Content translation/V2"
Reply to "Prenvented publishing because of "unmodified machine-translated text""
Reply to "Supported language pairs for MT"
Reply to "Handling of templates"
Reply to "In production, but many problems"
Reply to "Attribution required"
About this board
This is feedback page about the next version of Content translation tool.
Also, it will help us a lot if you will indicate the following things:
- Which article were you translating and to which language.
- Which browser did you use (Firefox, Chrome, Microsoft Internet Explorer, Safari, Opera, etc.)
- If you're getting any errors, please provide log from browser console if possible. To open browser console: press Ctrl+Shift+J (or Cmd+Shift+J on a Mac).
"Automatic translation failed!"
this tool is apparently broken
This article at en.wiki was apparently created using this tool. If you look at the wikisource for that article you will see that all of the en:Help:Citation Style 1 templates are not stored as templates but are stores as raw html. This is wrong. There is a similar problem described at phab:T209493 though that phab task is about copy/pasting named references using the abomination that is ve, but since the symptoms are similar, perhaps if the translation tools share ve code...
FYI here is a link to the details of the page creation (and to note, the 'creation' was actually as a userspace draft, that was later moved to the article):
File name prefix: pls unlimit adding "Wikipedia:" on jawiki
For CX2, I worked on a page with prefix Wikipedia: on enwiki. When I tried to name the ja page title as starting with Wikipedia:, the CX2 does not accept so that I ended up naming it The_Wikipedia_Library/OABOT. It has been moved by a user anyway to Wikipedia:OABOT.
How can I name the translated page with "Wikipedia:" prefix when using CX2? Or, can you think of adding a button so that published translation is filtered as it needs to be moved/renamed? Or, I am concerned that we owe local community editors think as me repeating same unwise action, as page names are something each community is very specific about I guess.
When you tried to create the page w:ja:Wikipedia:OABOT did it not create? It appears at https://ja.wikipedia.org/w/index.php?title=Wikipedia:OABOT&action=history and doesn't seem to have been moved? When (date/time) were you trying to do these tasks?
No, my original publication of CX2 output was not named w:ja:Wikipedia:OABOT since the system rejected the title; so I named ":w:ja:The Wikipedia Libray/OABOT" instead. Then later a user moved it to the expected namespace, hence it is now at :w:ja:Wikipedia:OABOT.
I'm not able to duplicate this. I was just able to create a page on jawiki from content translation:
2019-03-10T21:21:05 Xaosflux (talk | contribs) moved page Wikipedia:Sandbox3 to 利用者:Xaosflux/Sandbox3 (Test) (revert)
2019-03-10T21:20:17 Xaosflux (talk | contribs) created page Wikipedia:Sandbox3 (ページ「Indian River State College」の翻訳により作成) Tags: ContentTranslation, ContentTranslation2
Would you try CX2 with a Wikipedia: prefix page please?
I created it as "Wikipedia:Sandbox3" already, not sure what else to look at?
I mean, did you use CX2 and put some Wikipedia:xxx page as the translation source, on your left pane? Then start translation actually? Unfortunately, reating a blank page on jawiki is out of the focus on my query. If you had done with "Wikipedia:Sandbox3", my appology. How do you translate a Sandbox ?
I started with a normal article from enwiki, and translated it TO a Wikipedia: page on enwiki. I was unable to translate FROM pages other than articles. I had no problem at all creating the destination page via the CX tool in the Wikipedia: namespace.
You are right, normal pages can be renamed to Wikipedia:yyy name. I agree with you we are unable to translate FROM Wikipedia:xxx pages, a type of special page. Cheers,
Please look at the left of "公開する"(publish), you'll see a gear button. By pressing this, You can create sandbox page for you. So just make the translated article in your sandbox, create "Wikipedia: -" besides Contents translation, then copy the translated source and paste it to new page. In this case, don't forget history inheritance (履歴継承).
Case closed for me, and very thankful for your inputs; I wish your instruction will be seen by those not-so-techy translators like me. Backlist might hopefully get shorter faster, and new editors will be supplied with time tested wisdom Wikipedians had spent time to mediate “what to does”. Domo arigatow gozaimasu.
The bug which deletes reuse references was removed (now only in v2. v1 has the same problem even now). Thanks a lot!
I'm working on translation from English to Japanese. Since V1 I had a problem that the reuse reference style such as <ref name="name" /> disappears when I publish an article. Only the first defining refs survives. Actually English ref syntax and Japanese ref syntax are same so I'm fixing refs by copy & paste after the article is published, but this is too troublesome work. I saw the V2 solved some reference trouble so I tried it, then found that reuse refs just disappear before translating (when I touch the right field). Is there any way to reserve reuse refs?
Hum with Google Chrome I get this https://ja.wikipedia.org/wiki/%E5%88%A9%E7%94%A8%E8%80%85:Doc_James/Opioid_overdose
With Firefox it works better and I get this https://ja.wikipedia.org/wiki/%E5%88%A9%E7%94%A8%E8%80%85:Doc_James/Urinary_tract_infection
Would recommend that you use firefox at this point.
I tried firefox but nothing have changed...(please see the bottom of the article) https://ja.wikipedia.org/wiki/%E5%88%A9%E7%94%A8%E8%80%85:McYata/%E3%83%95%E3%82%B4%E3%83%BB%E3%82%B3%E3%82%A6%E3%82%A9%E3%83%B3%E3%82%BF%E3%82%A4/firefox
You will find many of refs have disappeared from original; https://en.wikipedia.org/wiki/Hugo_Ko%C5%82%C5%82%C4%85taj
Yes I tried aswell and got the same thing https://ja.wikipedia.org/wiki/%E5%88%A9%E7%94%A8%E8%80%85:Doc_James/Hugo_Ko%C5%82%C5%82%C4%85taj Looks like this needs fixing.
Using these examples, I created a ticket for the specific case of this style of reference. This fits in our immediate plans to improve references support, so we may be able to work on it in a few weeks.
I confirm that the problem have solved. Now I'm free from the troublesome works to fix refs manually. https://ja.wikipedia.org/wiki/%E5%88%A9%E7%94%A8%E8%80%85:McYata/%E3%83%95%E3%82%B4%E3%83%BB%E3%82%B3%E3%82%A6%E3%82%A9%E3%83%B3%E3%82%BF%E3%82%A4/test
Thank you very much!
Thanks for confirming the problem is solved, @McYata.
Please, keep using the new version of the tool and reporting any issue you may find!
Can this version be compatible with old versions like Content translation 1 ??? Cause months ago I tried to translate with CT/V2 and i can't publish an article, it's also is very slow. Thank you.
Version one and version two of the translation editor are not compatible with each other.
Prenvented publishing because of "unmodified machine-translated text"
When I tried to publish new translated article, the tool said "Your translation cannot be published because it contains too much unmodified machine-translated text". However it can't be, because first of all I chose "Use source text" instead of using "Automatic translation", which is contained in Content translation tool. Therefore the sentences which I translated contains 0% machine-translation.
Every time I translate article get warning about "unmodified machine-translated", and I have avoided it by choosing "Mark as resolved". Now this is the first time I cant find any way to avoid it. What can I do?
Thanks for the feedback, @McYata.
This issue has been fixed in the code recently, and the fix will be available to users during the next weeks. You can read more details in this ticket. If the error persists after a couple weeks, please let us know and we'll investigate the issue further.
I waited a few days and tried again. I could publish the article and haven't seen the same error until now. Thanks!
Well, I got the same error again...the first article the error emerged was "Gunpowder Empires", and this time "Serica". It might happen when I try to translate some articles which contains <blockquote>, which can't be rewritten in Contents Translation at present. (though I haven't confirmed it at all.)
Sometimes a different error appears, which says "detected spam contents. See "cs.com". " (I don't know the real original English message. Japanese message is 「あなたの翻訳にスパムの可能性のあるコンテンツを検出しました。詳細：.cs.com」.) What is cs.com? How can I deal with it?
Supported language pairs for MT
I was eager to see if the new machine translation services provide more support for German (e.g. to or from English). But I still get "Translation services not available for the selcted languages" (de->en). Should I be patient, or is there any config, tweak, cached pages that I overlook? (URL ends on "&version=2")
Thanks for the feedback, @Aeroid.
Machine translation support for English and German as target languages is not enabled yet. Support for new languages is expanded gradually, and I captured your comment in this ticket to keep track of the different requests in this area. Feedback from the community is very important to expand the support.
Meanwhile, an alternative could be to translate from German to Simple English, for which machine translation is available.
Handling of templates
Can we get CX2 to handle templates better? All the spacing goes missing when the template is brought over which makes it difficult to parse.
In production, but many problems
I see you have put V2 in production in frwiki for about a week. I checked the last 5 translations done with the new version, and not one is free of problems... I've reported the problems in Phabricator : T213255, T213256, T213257, T213258, T213259.
Has it been really tested before being put in production?
Thanks for the reports, @NicoV.
Version 2 is an opt-in feature that is available for Content translation users. We are making version 2 more visible for existing Content Translation users as it gets improved in order for them to transition from version 1 to version 2.
Version 2 uses the Visual Editor editing surface which provides a more solid experience editing wikitext compared to version 1. Version 2 is still in active development, and we have processes to test, get feedback and improve.
Given the diversity of the content, the number of languages supported and the many templates that exists on each wiki (that were not designed to be transferred across languages), it is not feasible to anticipate all the problematic combinations, and we appreciate the reports to identify such issues.
I know that similar issues about the quality of the wikitext produced were reported in the past for version 1. Those were hard to fix in version 1 without making a big architectural change. Version 2 represents such change, many of those have been already fixed, and it is in a better position to get more fixed.
@Pginer-WMF I do understand this should be an opt-in feature, but as it is still beta (and still generating a lot of problems in the results), why do you advertise it so much? For example, I've disabled CX in my preferences (beta tab), but I still get popups messages to use it whenever I end up on a page that was deleted because it's not admissible... Why don't you only give notifications to users that have explicitly requested to receive them, rather than forcing them on users who requested not to have them?
Also, why don't you give more prominent message to explain that CX and CX2 is "beta" software and that users opting in should check the result of their translation, especially if they create translations in the main namespace?
CX has been creating buggy articles in production for 3 or 4 years now, and it keeps doing so: when will something really be done about this?
And when you say "it is not feasible to anticipate all the problematic combinations": I would agree if the problems were on a small number of edits, but they appear on a vast majority of articles. For example, I just checked 1 translation, and I reported T213419...
Hi there! The content of Wikipedia is licensed under a CC BY-SA license which requires attribution (and sharing alike). This tool should provide attribution using templates made for this use, e.g. https://en.wikipedia.org/wiki/Template:Translated_page --Palosirkka (talk) 14:48, 31 December 2018 (UTC)
Users are required to order a en:Wikipedia:Requests for page importation, if not done by a bot.