[Feature Request] Post language selector in the compose view.

[stevenroose]

Many people post in two (or more) different language (English and their native language, mostly). Instead of a single "posting language", it would be nice to be able to select multiple ones and then get a drop-down selector between those in the compose view.

The "auto-detect language" feature doesn't seem to work in many cases. I personally get a lot of French and German posts of users telling me their setting is auto-detect. I only want to see their English posts, since I'm not fluent in French or German.

Implementation

1. Support multi-select of posting language in Settings.
2. In the compose view, next to the "CW" tag, have a tag with the most recently used language from the list in the settings. F.e. "en"
3. When clicked, show a select box with the other languages in settings.
4. The box in the compose view could of course be hidden for users with only one language selected in settings.

(relevant: #3201)

[stevenroose]

As pointed out here, it would be helpful if the language detection was not only done for long toots. The limit now seems to be 140 characters.

[trwnh]

Technically a duplicate of #8753 but that one was closed "to keep the clutter down"; obviously, I disagree and believe that a per-post language picker would be a useful addition for multi-language users.

[stevenroose]

@trwnh I also propose to have users that post in different languages explicitly ask for the "clutter" of having the language selector.
@Gargron would that be worth considering? It would be a real improvement for people following popular multi-cultural hashtags.

[stevenroose]

@trwnh @Gargron ping :)

[hope1]

I write and read Chinese, English, and Japanese. Now, I also have a browser Stylus override where I set different fonts for Chinese and Japanese, instead of the system default one.

Specifically I set serif (SongTi/Mincho, as they are called in local typographical traditions) fonts for the two languages, because they look better in longer text. In my opinion, text composed with Hanzi/Kanji are more readable with a serif font (SongTi/Mincho) than a sans-serif font (HeiTi/Gothic), and the ubiquity of the latter on screen is an unfortunate result of English hegemony and the better readability of sans-serif fonts in Latin script. With language selectors I could also set different override fonts for Chinese and Japanese, catering to glyph variation in the two languages, which is all very nice.

Now, this Stylus override works quite well on Twitter overall because it applies language detection to all posts, and because its language detection is quite reliable (admittedly I don't know about how all this applies to other languages that I don't speak). Even there I'd still like to have manual selection of post language mainly because that sometimes people (including me) would post a Chinese-written text that nonetheless uses Kana characters, which might be proper noun, or might be because it is about Japanese language itself. And these posts would be detected as Japanese, but I'd prefer them be treated as Chinese text using Chinese glyph variations.

Now language detection seemed to work less well on Mastodon, and it doesn't work at all for shorter posts. Everything would have worked out nicely if the user is given a good-faith choice to mark the language of their post. In particular, because people won't retrospectively mark the language of their post, this results in a loss of metadata information, useful to comprehension, because there wasn't an interface for it.

To an extent this is also an accessibility issue and just an aesthetic one: my Stylus workaround basically enables me to use a font with good readability specific to each language. Which font might be more readable regarding to one language might also change depending on which device the user is using: Chinese users more often than their English counterpart has to resort to a bitmap font on a smaller or older device. Yet another reason for ease of choice, as it relies on the community to properly mark outliers enough to complement auto-detection.

One can also think of other accessibility use cases for user-generated language marking. Say, a screen reader, which would be rather language-dependent and input-sensitive, and a disable user should be able to have their computer read out each message in the languages they understand in the language of that message, and skip the ones that they won't understand.

[melissaboiko]

I think not being able to mark the language of one's posts will probably undermine screen readers totally, including for the image captions? There's no point in captioning an image if the screen reader will garble it trying to read in the wrong language.

An UI to set the language that would set lang= in the HTML would make captions work for major screenreaders including JAWS, NVDA, TalkBack etc.

[stevenroose]

Amazing, I'm very happy with this feature! The next step for perfect multi-language UX is #11180! :)