#askFedi Does anybody know if there is an #ActivityPub software that *produces* multilingual objects leveraging the contentMap feature?
-
Relevant Mastodon issues:
https://github.com/mastodon/mastodon/issues/11013
https://github.com/mastodon/mastodon/issues/16692TBH, I find the comments suggesting that “DeepL [or any other client-side translation service] is enough” to be appalling.
Of course those who follow me know that I'm a big fan of #multilingual authoring if not else because of my 2023 article about implementing the #SMIL #switch element in #HTML plus #CSS —which isn't actually possible, requiring a little bit of #JavaScript too:
(And I'm not saying Mastodon should implement them using the trick above; it definitely needs a better interface.)
-
#askFedi Does anybody know if there is an #ActivityPub software that *produces* multilingual objects leveraging the contentMap feature?
@oblomov Post like https://no.fedimeteo.com/trondheim/p/1761455314.991738 might benefit from using multilingual objects
-
@oblomov Post like https://no.fedimeteo.com/trondheim/p/1761455314.991738 might benefit from using multilingual objects
@aslakr yeah, the recent conversation on the possibility to make the fedimeteo posts more international is exactly what triggered this thread in the first place 8-) Of course in that case the _production_ side is up to the platform used by fedimeteo, which is snac, but Mastodon being at least able to properly present such posts would be very useful.
-
#askFedi would you like to have the possibility to make multilingual posts, i.e. a *single* post/toot with the text available in multiple languages *at once*? Currently, you have to either do separate posts, or squeeze multiple languages within the character limit of your instance and only mark it for one of the languages (see e.g. https://sociale.network/@oblomov/110872023337803783 + https://sociale.network/@oblomov/114983566130691059).
@oblomov Why not have an option for machine translation of posts while reading?
-
@helge thank you very much, that's an interesting project. You're absolutely right about the issues about a multilingual posts, but I don't think they are unsolvable, and even a provisionally inadequate solution would be a better starting point than no support
For example: an indicator could be shown that other languages are available for a post, and allow switching to a different one (compare with the Edited feature), and users could choose the languages to show by priority, if missing, etcThanks! I will appreciate when people call funfedi.dev interesting.
Anyway, I opened an issue explaining more on how language support works.
Issues like this are why I submitted a proposal to nlnet to rework funfedi.dev.
I think working out a POC for multi language support in the Fediverse would be interesting. However, the first step here should be to find somebody making multilanguage social media posts, and ask them: "What features do you want? What's your workflow?"
-
#askFedi Does anybody know if there is an #ActivityPub software that *produces* multilingual objects leveraging the contentMap feature?
@oblomov 저는 워드프레스에서 contentMap에 접근하는 방식에 대해 이슈를 작성했습니다.
Do you want to include multiple language versions when publishing an article of more than 7,000 characters in WordPress?
If you include 14 language versions, a single `Create` activity would contain about **98,000 characters** of content — and every time you update one of those localized versions, the `Update` activities will keep accumulating.ActivityPub Internationalization · Issue #2257 · Automattic/wordpress-activitypub
When using contentMap, it is generally better not to include content, since the two are functionally redundant. The only real difference is that contentMap carries explicit language tags. A default language can be specified via the JSON-...
GitHub (github.com)
-
@oblomov 저는 워드프레스에서 contentMap에 접근하는 방식에 대해 이슈를 작성했습니다.
Do you want to include multiple language versions when publishing an article of more than 7,000 characters in WordPress?
If you include 14 language versions, a single `Create` activity would contain about **98,000 characters** of content — and every time you update one of those localized versions, the `Update` activities will keep accumulating.ActivityPub Internationalization · Issue #2257 · Automattic/wordpress-activitypub
When using contentMap, it is generally better not to include content, since the two are functionally redundant. The only real difference is that contentMap carries explicit language tags. A default language can be specified via the JSON-...
GitHub (github.com)
@oblomov What I mean is: microblogging isn’t Wikipedia, and traffic isn’t free.
Make active use of automatic translation instead. -
#askFedi would you like to have the possibility to make multilingual posts, i.e. a *single* post/toot with the text available in multiple languages *at once*? Currently, you have to either do separate posts, or squeeze multiple languages within the character limit of your instance and only mark it for one of the languages (see e.g. https://sociale.network/@oblomov/110872023337803783 + https://sociale.network/@oblomov/114983566130691059).
@oblomov Honestly, automatic translation has become so good that I doubt it‘s worth adding this complexity to Mastodon/ActivityPub. The UI for making posts already has too many option and buttons and stuff.
Translation can and should be a concern of the client. It works very well elsewhere.
-
@oblomov Why not have an option for machine translation of posts while reading?
@apicultor several clients already do that, and it can work as a fallback if the author doesn't provide a translation of their own, but what I'm talking about here is the possibility to author multilingual posts directly, which several people do already, but through impractical gimmicks like the ones I mentioned in the poll post.
-
@filobus hard disagree. Client-side translation should only be considerd a low-quality fallback for the cases where the author did not/could not provide a translation themselves. Allowing users to provide their own translations is an important step to improve the internationalization of the Fediverse.
@oblomov @filobus My guess is that only 0.1% of users would ever use this feature. But the additional UI for it would probably be forced upon everyone.
In many cases I also expect that the manual translation of a non-native speaker is worse than an automatic translation. I see this regularly happen here at work. -
@oblomov @filobus @valhalla @Puxi i don't want this feature because people will provide text with different meanings in different languages. Let's say put a witty clever take in English and blattant hate speech in French. So English reader will retoot and give visibility to hate in good faith.
It's just like alt text don't assume people will make good usage of a feature just because it exists
-
Thanks! I will appreciate when people call funfedi.dev interesting.
Anyway, I opened an issue explaining more on how language support works.
Issues like this are why I submitted a proposal to nlnet to rework funfedi.dev.
I think working out a POC for multi language support in the Fediverse would be interesting. However, the first step here should be to find somebody making multilanguage social media posts, and ask them: "What features do you want? What's your workflow?"
@helge @nlnet I wrote something up in my comment to the Mastodon issue https://github.com/mastodon/mastodon/issues/11013#issuecomment-3447834033
I'm not sure if it's sufficient as a starting point, though.
-
@oblomov What I mean is: microblogging isn’t Wikipedia, and traffic isn’t free.
Make active use of automatic translation instead.@thaumiel999 I responded here https://github.com/mastodon/mastodon/issues/11013#issuecomment-3448312779, but the short of it is:
* authors should be allowed to write multilingual content with proper metadata, regardless of the content form (blog, microblog, image description, video subtitles or anything else);
* even with extreme cases like the one you mention, a multilingual post is still going to be more lightweight than any multimedia content
* multilingual posts are more efficient than the current workarounds. -
@oblomov Honestly, automatic translation has become so good that I doubt it‘s worth adding this complexity to Mastodon/ActivityPub. The UI for making posts already has too many option and buttons and stuff.
Translation can and should be a concern of the client. It works very well elsewhere.
@bitbonk ActivityPub already has this feature (see contentMap etc, which Mastodon *already* uses, but only for a single language, unnecessarily duplicating content). And no automatic translation will ever be a good substitute of the author's own words.
The UI can be enabled only for those who want it. -
@oblomov @filobus @valhalla @Puxi i don't want this feature because people will provide text with different meanings in different languages. Let's say put a witty clever take in English and blattant hate speech in French. So English reader will retoot and give visibility to hate in good faith.
It's just like alt text don't assume people will make good usage of a feature just because it exists
@lutindiscret @filobus @valhalla @Puxi my poll currently has a 60%+ of people interested in the feature. The kind of abusive behavior you mention would be a bannable offence, dealt through moderation like any form of abusive use of the platform. And just like alt text, we should not reject a positive feature only because some may not make good use of it.
-
Of course those who follow me know that I'm a big fan of #multilingual authoring if not else because of my 2023 article about implementing the #SMIL #switch element in #HTML plus #CSS —which isn't actually possible, requiring a little bit of #JavaScript too:
(And I'm not saying Mastodon should implement them using the trick above; it definitely needs a better interface.)
Interesting, from the comments I'm seeing, this feature seems more controversial than I expected. There seems to be in particular a crowd that seriously believes client-side translations to be a superior alternative to the author's own words, rather than an “extrema ratio” fallback for untranslated content.
I'm starting to see why Google has been pushing that autotranslated crap on YouTube, the AI brainrot is already dramatically widespread.
-
@bitbonk ActivityPub already has this feature (see contentMap etc, which Mastodon *already* uses, but only for a single language, unnecessarily duplicating content). And no automatic translation will ever be a good substitute of the author's own words.
The UI can be enabled only for those who want it.@oblomov From the UX perspektive Mastdon is already quite … involved. Adding yet another options doesn’t make it better.
-
@oblomov From the UX perspektive Mastdon is already quite … involved. Adding yet another options doesn’t make it better.
@bitbonk any additional UI elements can be hidden behind a preference, so people that post monolingual can do so without seeing anything different, and people who wish to post multilingual can do so.
-
Interesting, from the comments I'm seeing, this feature seems more controversial than I expected. There seems to be in particular a crowd that seriously believes client-side translations to be a superior alternative to the author's own words, rather than an “extrema ratio” fallback for untranslated content.
I'm starting to see why Google has been pushing that autotranslated crap on YouTube, the AI brainrot is already dramatically widespread.
I can appreciate the worry about the potential of abuse from people hiding bannable offences in the translations, and how this could make work more difficult for moderators, but I don't share the pessimism: even now, if anything gets into the timeline in a language that isn't understood by the moderators, they're unable to make easily an informed decision. Those tools remain in place, and the report form can include (auto-filled) information about the language for which it is being sent.
-
@bitbonk any additional UI elements can be hidden behind a preference, so people that post monolingual can do so without seeing anything different, and people who wish to post multilingual can do so.
@oblomov Yes, I understand that. It is still added complexity for users, client implementers/designers and mastodon maintainers for (in my eyes) little benefit.
