Wikimedia Language engineering/Wikimania2015/Translathon1

Notes from Content Translation Workshop - Day 2 of Wikimania 2015

About the tool

Content Translation is a tool to help you create new articles missing in your language by translating them. It has been already used to translate more than 10K articles. It is a beta feature: you need to log-in at Wikipedia, access the beta link and activate it (for the different languages you want to translate to and from). More info at https://www.mediawiki.org/wiki/Content_translation

About this session

The goal is to try the tool together, create some content and share our experiences to make the tool better. This etherpad will be used for coordinating and list our achievements. Total time 1h: 5min. Intro and set-up. 35min. Translate all you can. 20min. Discussion about the tool.

Follow [1] for live updates. Remote participants are welcome, share the etherpad link with your friends: http://etherpad.wikimedia.org/p/translathon (retired and content copied over to this page)

Languages of the participants

List the languages you speak to see how much we can spread content across languages in this session:

en: English
es: Spanish
et: Estonian
de: Deutsch
it: Italiano
gu: Gujarati
ja: Japanese
nl: Nederlands
or: Odia
pa: Punjabi
ca: Catalan
id: Indonesian
eo: Esperanto
ru: Русский
uk: Українська
ur: Urdu
ne:Nepali
mai:Maithili
tlh: tlhlngan Hol
ro: Romanian
tyv: Tyvan

Articles published

When you publish a translation, paste the link below so that we can celebrate our achievements:

w:nl:William Morrison
w:ru:Дионисады
w:uk:Авакянц Сергій Йосипович
w:ca:URL Snooper
w:ca:Autobuses del Norte
w:id:Tamale
w:pa:ਚਾਪੂਲੀਨੇਸ (Chapulines)
w:ca:Corioamnionitis (medical article!)
w:nl:Paul Carlson
w:de:San Ignacio (Belize)
w:id:Metro Portales
w:et:Gyula Juhász
w:ne:ब्लेनहेम दरबार
w:gu:મેક્સિકો સિટી
w:gu:કેરમ
w:ro:CHAP
w:de:Banker Horse
w:or:ନ୍ୟାସନାଲ ମିଉଜିଅମ ଅଫ ଆନ୍ଥ୍ରୋପୋଲୋଜି (National Museum of Anthropology (Mexico))
w:ur:سروے پلی رادھاکرشنن (Sarvepalli Radhakrishnan)
w:id:Hotel Hilton Reforma Kota Meksiko
w:ja:天津租界

Articles to translate

Translate any article you want! But if you ran out of ideas when looking for articles to translate, here are some. When Content translation is enabled, the list of language links will show some missing languages you can translate to. Look at the published article list, some may be missing in a language you know.

Tool to find articles missing in a language: https://tools.wmflabs.org/not-in-the-other-language/
List of World Heritage sites: https://en.wikipedia.org/wiki/Lists_of_World_Heritage_Sites
Medical articles: https://en.wikipedia.org/wiki/Wikipedia:WikiProject_Medicine/Translation_task_force/RTT
Galician sites: https://en.wikipedia.org/wiki/Wikipedia:Galicia_20_-_20_Challenge
https://en.wikipedia.org/wiki/Tree_shaping

Questions?

Once I start editing, I lose the translation and therefore some of the trickier words

Issues found

Add quick notes on the issues you find while translating (if you want to elaborate feel free to comment on the project discussion page: https://www.mediawiki.org/wiki/Talk:Content_translation ).

Lack of sync is really painful. It seems that the tool will let you open an existing target article, and includes the old categories, but *no* article text, which I assume means it will quietly overwrite everything? If this feature isn't ready yet, the tool should lock you out of such an action.
Warn earlier in the workflow (during "new translation") that machine translation will not be available for your language pair.
In the list view, hovering over the ellipsis is fine, but clicking on the ellipsis should either have the same behavior as hover, or be a no-op. Instead, clicking on the "..." behaves like a click on the article.
VE bug: selecting a link and replacing its text inconsistently preserves or breaks the link.
Especially for long articles, it's going to be a huge win if the translation can be done collaboratively, in parts. There could be a checkbox to make your partial work public, it would be in the Drafts namespace, etc.
When marking a link as missing (red link), the target link should be modifiable+1
Wide tables are not fully visibly nor translatable
Typing (and loading) can be very slow
When visiting Special:CX via contribs menu, it immediately asks to create a new translation even if you wanted to continue existing one
Link text was not adapted while target was
Sentence highlighting not matching, showed too much on target side
The default target title should be populated automatically after the source title is entered.
"Existing article" validation should only be performed on the target title, never on the source title. If the target == source title, it will have the current behavior anyway.
Lost my cursor. Suddenly I can hover and click on things, but no cursor, typing does nothing. Arrow keys scroll the page. I'm getting this a lot, will try to track down steps to repeat. Seems to be right after I populate a new paragraph. Turns out this only happens with the paragraph I mention below, with the bogus "[2][2]" reference in the source text.
Refreshing the page after beginning to translate causes you to go back to the beginning of the workflow. Example URL: https://en.wikipedia.org/wiki/Special:ContentTranslation?page=Virginia+Gonz%C3%A1lez+Polo&from=es&to=en&targettitle=Virginia+Gonz%C3%A1lez+Polo. (second, related bug:) Clicking "start translation" in this case begins an empty translation, when it should resume editing at your saved draft. Happily, the in-progress work is still saved, so you can resume editing from the index listing.
No reason to collapse the link infoboxes in the right sidebar when there are only two, and they would fit on the page expanded.
There is no "undo" after you "clear paragraph". Autosave screws you in this case.
It's possible to defeat the autosave. If the page isn't saved, we should get a warning in the "beforeunload" hook before closing the window.
Show a spinner while target content is loading.
Citation numbering is really screwy. I have [2][2] as the first two citations in a source paragraph (in this case, should just be [2]). In the machine translation, an unlinked "[2]" appears at the point of citation, but an additional linked [2][2] appears at the end of the sentence.
Selecting text causes bizarre things to appear in the right sidebar.
"Restore" should warn before overwriting changes.
Doesn't translate or copy external links. No apparent way to add these manually.
No way to add categories.
Arrow keys should be able to move the cursor between paragraphs (nice to have).
External links section content was not displayed in the source text.
References were badly corrupted after being published. Looks like it was unable to map the "Template:Cita publicación" template to enwiki's.
When the template is mapped, template parameter names are not.
Grey missing languages were not appearing, maybe because of monobook.

What worked well

Machine translation is luxurious کہ .
VE is very convenient.
Autosave works nicely.
"Publish" works well.
Triple-click selects the paragraph, as you would expect.
Moving text containing links does the right thing.
The design is beautiful and usable.

What can be improved

I'm missing a button in the toolbar to add internal links to the translationp
Links in the original version don't work (they point to the target language version)
Machine translation from English into German (not working) +1
Insert links manually (in case inserting them from the menu doesn't work)
Hovering on a word in the source text should give you a synonym popup.
When clicking a link at the translated contents,
Accept and integrate Wikitext
Machine translation from Hindi to Urdu (not working)