Xanthor - xanthor@occitanie-decroissance.info

Xanthor

xanthor@occitanie-decroissance.info

Xanthor

Sun, 27 Jan 2019 22:33:44 +0100

Hello #Fediverse

What is the good practice to specify language of posts?
Hashtags, metadata, categories, something else?

!Hubzilla Users Support @Mastodon @Socialhome HQ @Osada ☸ Announcements @diaspora* HQ @pixelfed @Friendica Support #federation

show all 10 comments

Giacomino

Sun, 27 Jan 2019 23:10:43 +0100

sorry, but is not my cup of tea.

Giacomino

Sun, 27 Jan 2019 23:49:10 +0100

...but I also think that depends on the software/network, since they are developed differently.

Megver83

Mon, 28 Jan 2019 01:17:39 +0100 from Diaspora

IMO categories and/or hashtags

King Emir

Tue, 29 Jan 2019 21:13:33 +0100

There are people in the fediverse who follow the hashtags #français or #deutsch for example. If you want to reach them, the use of these hashtags makes sense. They can be combined with following tools.

As I understand it, hubzilla categories are not federated to the world outside of zot. But their usage still makes sense: I find categories helpful to organize posts on my own hubzilla channel or on a hubzilla forum. To have a neat and well organised multilingual news stream for my web identity, I use categories on my hubzilla channel. The posts can be filtered for languages by the observer by using the links to the categories.
Thus, for text-dominated topics I use separate channels for each language, for example

feminisme
Feminismus
environnement
Umwelt
...

But for visual/audio/video-focused topics one multilingual category is enough

Musik / musica / musique / music / 音楽
Bildende Kunst / Visual arts / 視覺藝術
Photo 写真

Another method is to have multiple single-lingual news streams: Create a channel _fr@ and a channel _de@ for example, and give yourself (@) the permission to post on both walls of these new channels. This method allows your connections to follow the you (or a topic) in only one language.

I don't know much about metadata. The only implementation I used so far is the bbcode environment [observer.language=en][/observer] which defines metadata not of the post, but of parts of the post. I stated my concerns about its usage and some ideas how to deal with them here:

King Emir wrote the following post Wed, 28 Nov 2018 21:17:57 +0100

Allowing language switching by observer for each multilingual content in Hubzilla

Language switching performed by observer

Websites (see hubzilla.org for examples) and the help pages in #Hubzilla provide a language switcher (with a dropdown menu) per page. This allows the visitor/observer to choose between the language versions of the content for each content individually. The observer is empowered (selbstbestimmt, souverän, ...) in these cases.
I would say this mechanism is the state of the art in web publishing, with langswitcher in Grav, wpglobus in Wordpress, the translate extension in MediaWiki and so on allowing it.
I propose to use this language switcher or something similar also for the rest of the publication types which exist in Hubzilla (e.g. wiki pages, articles, posts, cards, comments ...). This tool should allow the observer to see in which languages the content is provided and then to switch between these content versions. For that purpose, the language of the content could maybe be defined as customary by [observer.language=...][/observer]. See conceptual visualisation below, but I can also imagine a collapse/expand mechanism as done by the [summary][/summary] tag, maybe better realizable since some content has no header (e.g. comments).

Language switching performed by the system (not discernable by the observer)

Hubzilla recognizes #multilingual content in wiki pages, articles, posts, cards, comments and so on when defined by [observer.language=...][/observer]. It then only shows (respectively hides when defined with an !) the content corresponding to the observer language¹. While I agree with showing the content like this as the default content, I don't agree with the fact that the other versions of the content are not visible. The observer is not aware that there are other versions of the content in other languages. Because of this automated hiding mechanism the observer is subject to the author's authority which is irresponsible from an ethical point of view. The usage of this bbcode tag in content which is accessible by more than one observer can create confusion. I used this bbcode tag once in a comment to a forum thread in which the discussion was having comments in 2 languages (here), and I feel bad having done so (there was no confusion created, but still I feel having cheated on the readers). Hiding/showing content should only be done by using the Access Control List.
¹ The observer language is set by the browser. This setting can be overwritten by the observer by using the Hubzilla app /lang which is accessed on its own path. This app is thus not usable for language switching while reading content.

Please vote

Yes: I like that idea of allowing observers to recognize all multilingual content as such and to switch the language for each multilingual content.

No: I don't like it, language switching for [observer.language=..][/observer]-type multilingual content in Hubzilla should be done only by the system.
!Hubzilla Development

https://hub.disroot.org/channel/emir/&mid=c43573e93c0fb4a327c44e656ff21a74bff18183c6cc8861c0c533b94d868956@hub.disroot.org

King Emir

Tue, 29 Jan 2019 21:22:59 +0100

when it's not about posting, but about reading posts, then in Hubzilla you can filter the contacts for languages, as mentioned here: #^https://hub.farthinghalearms.com/display/fefdda62bb906dbb463d8fb8d89d3a0344597da460a30ee713497764e761543a@hub.farthinghalearms.com&owt=cb28f72a3849c36c9a763c33f1a0b8b4

Olaf Wendt (Hz)

Tue, 29 Jan 2019 21:35:46 +0100

@King Emir Making use of two separate channels, fed by one "central instance" channel is a fantastic idea. I could accept new connections and let them observe the channel most suitable for them. Ultimately, the central instance could even be hidden for public.
If observers are multilingual they would still be free to connect to my other channel(s). I will definitely think about how to exactly implement this, because so far I'm not very good in adding categories to each post I write. Thank you for the input!

Xanthor

Tue, 29 Jan 2019 22:07:09 +0100

@King Emir
My concern is not really to reach language-specific readers, but to allow people to filter (and exclude languages they don't read, if they want to). I also intend to be compatible with full #fediverse, so Hubzilla categories, yet really useful, can't be the solution.
My web developer background suggests to use metadata and the observer.language you talk about seams promising.
Hubzilla could add an UI to switch this on a post basis, and set a default value per channel.
My main question is: is it #activitystreams compatible? I mean, is that metadata is available in an ActivityStream description? If it's true, then each fediverse software could choose how to handle it and inform the reader.

I'm not found of multi-channel, since there are other reasons to multiply channels, and if we add the language dimension, the number increases polynomially ;¬)

I know the language-filter feature of connfilter (even if for some unknown reason this feature is disabled on my hub), and I wanted to know how does it work? Does it use metadata, or language recognition?

Thanks for answers! (!Hubzilla Development )

Mike Macgirvin

Tue, 29 Jan 2019 23:36:46 +0100

Connfilter works by language recognition. It is reasonably accurate, but there is a minimum length threshold, by default 128 characters. Below this threshold the accuracy gets pretty dodgy so it is always allowed. In Zap and Osada, a similar but upgraded tool exists called 'Content Filter'. The major difference is that this includes the original connfilter functionality but also provides a global filter so you can apply filtering to your entire stream instead of one connection at a time. Since content is classified and filtered by the storage engine, the transmission protocol is not relevant.

Xanthor

Wed, 30 Jan 2019 19:43:57 +0100

I just checked and as I expected, ActivityStreams has a way to specify language of content transferred.

Multilingual example:

{
  "@context": "https://www.w3.org/ns/activitystreams",
  "summary": "A simple note",
  "type": "Note",
  "contentMap": {
    "en": "A <em>simple</em> note",
    "es": "Una nota <em>sencilla</em>",
    "zh-Hans": "一段<em>简单的</em>笔记"
  }
}

By the way, according to the spec:

Activity Streams 2.0 publishers SHOULD explicitly mark the language of natural language properties if they are known, using either map properties or a default language tag.