Proposal talk:Add Translate extension to Wiki

This page might get busy so it's recommend enabling Convenient Discussions in your gadgets to make reading and replying easier.

Handling non-English content

All translations are based only on the English source. You cannot edit translated versions of pages.

I do not understand how non-English content without any translations is handled? Do you just keep the old page? What if it has a language prefix? --Tigerfell  (Let's talk) 19:58, 29 July 2021 (UTC)

@Tigerfell: Yeah I forgot to account for this. You can use a different source language for translated pages. I would say that whatever is decided as the source language, if that page has a prefix, it should be moved to a page of the name without the prefix when the translation of that page without the prefix's translation is complete (or if there is no translation, this can be done immediately). --Lectrician1 (talk) 21:55, 31 July 2021 (UTC)

The point is that not every information in this wiki is acutally available in any language. Additionally, some pages do not need any translation like France/OSM-FR/CA 2019-02-03. --Tigerfell  (Let's talk) 18:43, 13 August 2021 (UTC)

To answer my own question, one would probably move a page like this from the current location to some subpage .../fr and not mark this page as ready for a translation but that is mainly a policy issue (how the wiki community wants to handle such cases uniformly) and not a technical matter. --Tigerfell  (Let's talk) 21:21, 22 March 2022 (UTC)

Contributing without English language skills

All translations are based only on the English source. You cannot edit translated versions of pages.

When translated versions of pages are not editable, this excludes users without English language skills from contributing. Whatever Pros might exist for the Translate extension, this single point alone makes its use completely unacceptable for the OSM Wiki in my opinion. --Lyx (talk) 08:35, 28 March 2022 (UTC)

I agree with that sentiment, however I think this could be solved by allowing anyone, at any time, to sever the "this page is a translation of X" connection and instead continue working on the page as a stand-alone language version. That way, people without knowledge of English would not be excluded. Someone who would prefer to re-establish the link "this page is a translation of X" would then have to ensure that the English "source" is properly updated with the content added to the formerly "translation" page. Of course this is only possible if language pages are allowed to deviate from the "original" but that is a requirement anyway, see my issue below. --Woodpeck (talk) 09:52, 28 March 2022 (UTC)

Yes, this sort of system would still be acceptable. We are not forcing people to use the Translate extension if they don't want to. They can maintain deviance in content by still using the old translation system pages. Lectrician1 (talk) 18:17, 28 March 2022 (UTC)

I am not (only) talking about users making language- or region-dependent changes to documentation in other languages than English. OSM tagging is constantly evolving, and documentation in the Wiki plays a role in shaping that evolution. In my opinion it is important that all users can participate in that development regardless of which language(s) they speak. Because of that I believe we need to enable users to extend existing documentation, and write new documentation, in their own language; of course other users would still be encouraged to translate these additions into other languages, especially into English. --Lyx (talk) 00:45, 29 March 2022 (UTC)

Okay, and they can still do that. They would just need to translate their contribution into English first. Consistency in documentation needs to be rooted somewhere or else we'll end up with one big mess of conflicting tagging documentation, which is what we have right now in addition to things not being completely translated. Lectrician1 (talk) 00:51, 29 March 2022 (UTC)

Please remember that we are talking about users that do not speak English. So if those users want to add to the wiki, instead of adding it right then, they first have to find someone who has sufficient language skills in both English and one language of the user wanting to contribute, explain to that volunteer what they want to change and hope that the translator they found adds the translation to the English version before they can add their own original contribution as a translation. I would expect this process to take days at best and more commonly take weeks or months. If I where in the place of that user, I wouldn't bother and simply give up on contributing. --Lyx (talk) 01:18, 29 March 2022 (UTC)

I wouldn't think so. People would see the "This is a translated version of <source page name>". They would then click on the page and see it is in English. They could then attempt to edit the source and add their machine-translated English contribution in the appropriate spot since they could recognize the page layout pretty easily. Even if it's not grammatically correct, someone else could fix it.

They could also always start a discussion on the translated or source page's Discussion page and others could translate what they are asking for to change.

I also explained that language communities on OSM should set up their own documentation as to how to go about translating content to English and who can contact in what channels in their community. Right now this documentation doesn't exist since translating on this wiki is a free-for-all. I do believe such pipelines can develop and successfully enable contributions by all, especially since there are already many English speakers in the many of the different language Contact channels. Lectrician1 (talk) 01:34, 29 March 2022 (UTC)

Good point. There are users here who translate from Spanish to Catalan and they don't know English.

I also know Polish users who are expanding documentation in Polish based on their experience in mapping, but who don't speak English well. They are expanding it for other Polish-language editors. maro21 22:07, 29 March 2022 (UTC)

Conversation resources

If I understand it correctly, all translations need to be re-done, right? Looking at German as the most common language behind English in this wiki (Note: Polish might have surpassed it), we are talking of at least 2000~pages starting with "DE:Key:" and "DE:Tag:" only. Is there a way to import existing translations to Translate Extension? --Tigerfell  (Let's talk) 20:14, 29 July 2021 (UTC)

Yeah they all do need to be redone because of formatting and sync issues. There is no way to directly import. --Lectrician1 (talk) 21:55, 31 July 2021 (UTC)

I was just asking because I do not think that we have enough users to transition this manually. --Tigerfell  (Let's talk) 18:45, 13 August 2021 (UTC)

User rights

I would suggest to assign the "translate" right to signed-in users because non-logged-in users can not edit the wiki. I doubt that you can actually translate content if you do not have the "edit" right.

MediaWiki seems to have a user group called "Translation Administrators". Do we need this, too? --Tigerfell  (Let's talk) 20:19, 29 July 2021 (UTC)

I forgot users who aren't signed in can't edit. This is fixed.

About Translation Administrators, maybe.

I just gave the translate-messagereview and translate-manage permissions which the Translate Administrators have to the autoconfirmmed users. translate-messagereview allows for users to review translations, which can help determine the quality of translations. It also makes sure people do not self-review their translations (which they could do if they had this permission).

I'm not really sure exactly what translate-manage does. It says it allows "Allows users to update and manage message groups with web interface.".

Translation Administrators are given one role which is key to translation management called pagetranslation. Every time the source page is changed, the translations are going to have to be updated and this involves "remarking the page for translation" (this of course is to prevent vandalism of becoming present on both the source and translation pages or just mis-formatting causing the extension to fail at create accurate messages). Translation Administrators are going to have to be users who are at least semi-active and remark the pages. That, or we need a lot of them.

From what I see, we should probably have translation managers with the same permissions Mediawiki has. --Lectrician1 (talk) 21:55, 31 July 2021 (UTC)

Does it work in templates?

(I am eagerly wishing for an improvement of the system for translating articles and hope that this proposal hasn't died yet.)

> "clutter the English text with various markers"

In the last discussion on the topic which reverted the 7 year old decision to introduce this extension, some contributors deemed the use of markers as required by this extension as "unacceptable and [a] no go". Nevertheless there exist masterpieces like Template:Generic:Map Features:sport. Is it possible to use the extension in templates? Such map feature overviews might be a good start for rolling it out since typical users don't have to edit them very often and the current status quo is a horrible mess anyway. --Nw520 (talk) 20:46, 10 March 2022 (UTC)

@Nw520 Yes! Please see Template:Translatable template. For an example page it is being used on, see: https://meta.wikimedia.org/wiki/Main_Page/en. Lectrician1 (talk) 14:10, 15 March 2022 (UTC)

Content deviating on purpose

On many pages we have content in a language-specific page that is not present in the "original" English page, and this is done on purpose because the situation in the target are of the language-specific page is different from the "generic" English page. In other words, many pages are more than just a translation and this is not a shortcoming but a strength. How would this be modeled in the future? --Woodpeck (talk) 08:08, 28 March 2022 (UTC)

Is it valid for a translated page to have different content? The use of tags can differ per region (which would warrant a note in all translations or a separate region-specific page), but surely not depending on the language the mapper reads documentation in? Translations of generic pages having a target region seems like a bug. It's to be expected that region-specific pages often exist only in one language of course. --JeroenHoek (talk) 10:12, 28 March 2022 (UTC)

"Is it valid for a translated page to have different content" Yes. For example

• there are many local projects mapping something specific (for example AED mapping campaign in Poland) which deserve to be linked in language-specific entries
• https://wiki.openstreetmap.org/wiki/Pl:Key:amenity - discusses confusion caused by fact that Polish word "bar" and English word "bar" have distinct meaning (in Polish "bar" means "fast food" or sometimes "cheap low quality restaurant")
• https://wiki.openstreetmap.org/wiki/Pl:Key:bicycle_parking mentions Polish slang name for wall_loops
• https://wiki.openstreetmap.org/wiki/Pl:Tag:amenity%3Dschool mentions Polish-specific data source of high importance
• Pl:Tag:amenity=parcel_locker discusses naming issue specific to Polish language and local legal issue (Polish name "paczkomat" for parcel locker was registered as trademark with nonsense court decision upholding it). This page also contains Poland-specific tagging examples.

I would be able to list at least one example for nearly all larger pages in Polish. Pages should describe the same tag meaning, but there are many region-specific issues - and each region will have some issues of higher and lower importance. And ideally examples would use local context (this is trickier for languages used in many places across world like English or Spanish) Mateusz Konieczny (talk) 10:31, 28 March 2022 (UTC)

Most of your examples are Poland-specific, not Polish-specific though. The language related ones (like the 'bar'-example) are valid; does the proposed translation method block these? Adding a language-specific note to a translated paragraph should still be possible. --JeroenHoek (talk) 11:04, 28 March 2022 (UTC)

Many languages, including Polish, are nearly completely region-specific, often country specific. It makes sense to mention Poland specific things in Polish specific translation. Mateusz Konieczny (talk) 11:16, 28 March 2022 (UTC)

@Mateusz Konieczny In this case, making a section or subpage of Pl:Key:amenity for Polish-related uses of the tag at Pl:Key:amenity/Polska might be appropriate. That way the general documentation about how to use the tag is similar for all languages but in regards to usage of it in Poland, the documentation is specific for there. I could make or modify a template that links such a page to the main tag page. Lectrician1 (talk) 18:22, 28 March 2022 (UTC)

So documentation about Polish-language related aspects of amenity=bar would be on a separate page, right? Mateusz Konieczny (talk) 19:33, 28 March 2022 (UTC)

Correct.

Though we should probably try not do this. What if non-Polish speakers want to contribute amenity=bar to Poland? We shouldn't seclude mapping practices to those who only speak the language of that region. Including the Polish usage of the tag as a section on the original page or providing translations of the Polish page would be the best solution to this problem. Lectrician1 (talk) 19:40, 28 March 2022 (UTC)

Is it really a good idea to mention Poland/Polish-specific naming issue on Key:amenity listing and in translation into every single language? I definitely would not want to have on English (or Polish) page such discussion about false friends in German, Norwegian Bokmål, Urdu, Amharic etc Mateusz Konieczny (talk) 21:25, 28 March 2022 (UTC)

@Mateusz Konieczny: I wonder, would a Polish-speaking software developer encounter more difficulty understanding how the rest of the world uses a tag if the Polish-language documentation focuses on tagging practices in Poland? Or would that audience be expected to turn to something like TagDoc for a more data consumer–oriented view on tagging? – Minh Nguyễn ^💬 20:58, 28 March 2022 (UTC)

(1) Polish documentation is giving such info in addition to standard basic info, not instead of it (2) part of that is flavor (for example using locally known brands in usage examples) (3) Any decent Polish-speaking software developer will be capable of using English language documentation, even if their spoken English is bad. Software documentation is only extremely rarely translated to Polish (using Polish variable names rather than English is sometimes happening but is considered as a sign of inexperience or very bad and poorly qualified programmer, at least among people I know) Mateusz Konieczny (talk) 21:18, 28 March 2022 (UTC)

@Mateusz Konieczny: Thanks for the context. I guess I'm used to the not-so-pragmatic approach in Wikimedia-land of translating as much as possible, even the API errors. Local color shouldn't be a problem as long as we're smart about choosing translation units on the source page. In my years of translating into Vietnamese at Translatewiki.net (the original Translate-powered wiki), I found countless opportunities to add Vietnamese-specific turns of phrase or Vietnam-specific examples. For example, any time a message referred to a deadline in UTC, I would add a gloss in UTC+7 for users in Vietnam.

I would recommend choosing translation units that encompass whole paragraphs. Many of the pages on this wiki that use {{LangSwitch}} are really, really bad about itty-bitty translation units that don't fit well together. It's a tradeoff: longer translation units means more opportunities for outdated translations. But Translate is flexible enough that we can choose longer translation units for some messages that are likely to be more stable and shorter ones for those that could change frequently. Projects that use Transifex, such as Vespucci and iD, are basically doing the same thing already.

 – Minh Nguyễn ^💬 22:25, 28 March 2022 (UTC)

"does the proposed translation method block these". Apparently yes, it even lists "1:1 translations of pages" as primary benefit. I consider it as a primary problem and blocker Mateusz Konieczny (talk) 11:16, 28 March 2022 (UTC)

@Mateusz Konieczny Feel free to modify the Cons and Analysis sections to reflect this if you would like. Lectrician1 (talk) 18:14, 28 March 2022 (UTC)

@Woodpeck @JeroenHoek @Mateusz Konieczny Translated pages can only be based off of the original page in the original language. If we want to make pages for similar content that have content in different languages, then you would need to make a seperate page. Using you cannot translate and modify content from the same page.

For example, if we want Pl:Key:amenity to purposfully have different content than the source English page, then we should keep that old translation system Polish page and not convert it to a new-system page that uses the Translate extension. Lectrician1 (talk) 18:13, 28 March 2022 (UTC)

We should move content that is specific to certain areas and regions into paragraphs where it is highlighted that this is region specific content. Mappers from these areas which do not speak English should be put in a position where they can inform themselves about the meaning and usage of tags outside of their region (i.e. generic page translated in their language), and ideally we would translate this specific content also in English and other languages (marking it as region specific). This is also already done, albeit not to a great extent. —Dieterdreist (talk) 20:35, 28 March 2022 (UTC)

Comment from experience on other wikis

For people who have edited wikis using this extension such as Wikimedia Commons, Wikidata or the WMF Metawiki: What is your opinion of the extension and how have you found it? --Andrew (talk) 16:46, 28 March 2022 (UTC)

Is https://www.mediawiki.org/wiki/MediaWiki also using the same or something different? (my experience with it was horrifically bad and scared me away from editing) Mateusz Konieczny (talk) 19:36, 28 March 2022 (UTC)

@Mateusz Konieczny It is the same. Lectrician1 (talk) 18:12, 29 March 2022 (UTC)

I have three years of experience in translating OSM website on Translatewiki using this extension. The main difference is that these are translations for short messages - one word, one sentence - that are used on osm.org. Apart from adding translations for new messages I was mainly correcting other peoples' (random people not invloved in OSM) translations which were without context. There were a lot of errors due to the fact that random translators couldn't see where a particular message was showing on the website. It is translated in small portions, mostly single sentences without context. The context must be found by the translator. Unfortunately it took me several years to catch the fatal translation errors resulting from not knowing the context (i.e. where the message appears). These messages were translated by volunteers, often not using OSM at all. So let's say 2000 messages were translated by 100 different people, just because they know the language in question. In addition to translations errors, there were inconsistencies, i.e., the same term was translated differently by different people. This extension doesn't provide the context in which the sentence or word is translated, resulting in frequent translation errors. However, it is possible to add documentation (with hints where it is used) to each message, but that is extra work and an extra page to create.

Let's say that there is be a word "Argentinian" to translate into Polish using this extenstion. There will be probably many Polish speakers who know at least basic English and they'll translate it to "argentyński". But depending on context and place where is it used it can be:

• argentyński
• argentyńska
• argentyńskie [+inflected forms: argentyńskich, argentyńskiego, argentyńskiemu, argentyńskim, argentyńskimi, argentyńscy, argentyńsku, argentyńską, argentyńskiej + variants with capital letters in case they start a sentence (you don't know it without context because the English word is always capitalised)]
• Argentyńczyk [+inflected forms: Argentyńczykach, Argentyńczykami, Argentyńczykom, Argentyńczyka, Argentyńczykiem, Argentyńczykowi, Argentyńczyku, Argentyńczyków, Argentyńczycy]
• Argentynka [+inflected forms: Argentynce, Argentynką, Argentynkę, Argentynki, Argentynko, Argentynkach, Argentynkami, Argentynkom, Argentynek]

The example from last month [1] where I found an error that had been there for 13 years. Simple translation, "relation" in English is "relacja" in Polish. But in this case this word was using in a sentence here: https://www.openstreetmap.org/relation/0 where there is "Niestety, nie odnaleziono relacji #0." These are the consequences of translating separate words or sentences instead of the whole article at once.

This extension is better suited for software translations and short pages that will not change often. Our tags documentation changes and grows frequently.

After enabling this extension here I can see a lot of new users, "volunteers" who would like to "help" by translating only the easiest and the shortest messages. maro21 22:36, 29 March 2022 (UTC)

Example of page prepared for translation

• After a page has been marked for translation, using the VisualEditor for editing is not possible. This is a downside for new editors who don't know how to edit using Wikitext.
• The auto-generated translation message dividers (<!-- T:1 -->) can get in the way of editing.

This combo may cause major problems. Can you show how Railways would look after such necessary conversion? I suspect that

• It would become uneditable
• Such conversion would take enormous effort

Mateusz Konieczny (talk) 19:44, 28 March 2022 (UTC)

Harmonize languages first

Given that it is mandatory to have 1:1 matching pages, I propose to, before enabling such extension, to ensure that all content in other languages is present in English.

I suspect that it is impossible to achieve and in many cases harmful (which is my reason why I am suspicious about this extension).

But if that is possible and achievable and not harming wiki - then it should be done as the first stage. Mateusz Konieczny (talk) 19:46, 28 March 2022 (UTC)

See step 2 of conversion of typical pages. Lectrician1 (talk) 20:15, 28 March 2022 (UTC)

I know. My proposal is to start from it as I expect that this step will completely fail. And this way we will not end with deployed confusing extension that should be ignored by editors Mateusz Konieczny (talk) 21:20, 28 March 2022 (UTC)

@Mateusz Konieczny: Yurik attempted to identify some of the more structural discrepancies between languages when populating the data items some time ago. Many of these discrepancies remain in data items and can be identified by (intentionally bizarre-sounding) limited to language (P26) qualifiers. The idea is that they'd be replaced by limited to region (P48) qualifiers once both the data items and associated wiki pages are cleaned up. Obviously this audit didn't catch the less machine-readable nuances, but it's a start. – Minh Nguyễn ^💬 21:03, 28 March 2022 (UTC)

"Tagging documentation pages rarely change"

There is a claim "Tagging documentation pages (which are the primary use of this wiki) rarely change, meaning that we really aren't discouraging many editors from making contributions by limiting editing to Wikitext."

I consider this claim as false and highly dubious. What is the basic of claim that making much harder to edit Wiki pages in English the harm will be small? And even if edits are done rarely - how does "we really aren't discouraging many editors" follow? Mateusz Konieczny (talk) 19:55, 28 March 2022 (UTC)

Okay, maybe it's a bad claim. Removed. Lectrician1 (talk) 20:13, 28 March 2022 (UTC)

Feedback from translators

The proposal page is missing listing of places where you notified translators (you notified translators by posting also to non-English channels, right?).

If you have not notified translators who would be heavily affected by this changes you should use https://wiki.openstreetmap.org/wiki/Contact_channels listing to gather feedback and invite them to comment. Mateusz Konieczny (talk) 10:38, 29 March 2022 (UTC)

@Mateusz Konieczny Done in "External discussions" section. Lectrician1 (talk) 18:12, 29 March 2022 (UTC)

Thanks! At the very last Polish forum (maybe other languages) are also active Mateusz Konieczny (talk) 20:44, 29 March 2022 (UTC)

By "notify translators", I understand that people who are actively translating the Wiki (or were in the past) should know about this RFC, not people who are not involved in the Wiki. I don't know why the translators' work environment should be decided by people not involved in it. maro21 22:20, 29 March 2022 (UTC)

section "Working example in practice"

I added section Working example in practice where one can see how the extension works. So in this example every one can simply just translate a message "Thank you" without translating the rest of the text. So there will be situations that someone just translate "thank you" into one hundred languages and we will have a long list of languages. maro21 22:12, 29 March 2022 (UTC)