Posted under Bugs & Features
Now that we automatically bring in commentary on new uploads (which is a good improvement IMHO), we are also pulling in the default "無題" for untitled posts (which I'd normally not bring over). Should those posts be auto-translated to "Untitled" since they are so common? It's probably not worth adding the commentary tag to these though.
From a quick regex search there's about 4k commentaries with exactly "無題" in the title. A little under half of those also don't have a commentary description, making "無題" the entire commentary.
And now it seems @RaisingK went ahead and nuked all the posts with 無題 or "Untitled" in the commentary without telling anyone.
Just... Why?
He even nuked partials like in post #3751521, so now we have half the commentary for some posts.
And not just "untitled"; also 練習 ("practice") was nuked.
Updated
I overlooked this topic and nuked { "无题", "無題", "Untitled", "rkgk", "練習", "Practice" } under the assumption that such common, generic strings didn't qualify as "interesting additional information" (and users might mistake them for such where untranslated). I shouldn't have acted unilaterally, though.
...would you prefer I add them all back?
RaisingK said:
I overlooked this topic and nuked
{ "无题", "無題", "Untitled", "rkgk", "練習", "Practice" }under the assumption that such common, generic strings didn't qualify as "interesting additional information" (and users might mistake them for such where untranslated). I shouldn't have acted unilaterally, though....would you prefer I add them all back?
I believe that would be best until a decision can be reached.
Especially considering you did so without saying anything.
That being said, I like the idea of removing 'Untitled' commentaries, I've asked for something similar in forum #165278 and this would be another step in that direction. (I haven't thought about 'rkgk' and such but I don't really mind either way)
I think it would make both the commentary tag and the commentary request tag more useful, it feels weird to find a commentary request that I don't want to translate cause it's just untitled.
I don't know if there's any reason to leave them but they make the commentary tag less useful as is.
If anything is done about generic commentary, I think it should be automated, so users who don't know what 無題 and similar terms mean don't think it's some kind of oversight where the uploader just deleted or didn't add the commentary.
This is also an argument for keeping it in some fashion (perhaps also with automated translation as Shinjidude suggested); if it just gets silently erased, they might think it's a Danbooru bug and ask about it here repeatedly.
DreamFromTheLayer said:
I think it would make both the commentary tag and the commentary request tag more useful, it feels weird to find a commentary request that I don't want to translate cause it's just untitled.
I don't know if there's any reason to leave them but they make the commentary tag less useful as is.
This is the reason I was saying these simple auto-translates should probably not include the commentary tag if that's all there is to them. I agree that there isn't much meaning to these, and they could easily swamp commentary and commentary_request.
I'm ambivalent on keeping vs. losing them, but maybe somewhat closer on the "keep" side (which tends to be my default position on most things). If we nuke them, people could see that as confusing, especially if the title gets nuked, but the body doesn't because it has content. I also historically wouldn't bring these across manually if I saw them, but I do like the idea we've gone towards with automatically bringing commentary over. If we do, it might make sense to remain consistent and bring over anything, auto-translating the default and common fixed expressions.
"無題" in particular I think it would be a good idea to handle automatically. I'm not sure how far we should go with it though since "rkgk" is definitely user added, not system added. It's also something that is often part of a bigger phrase, where I definitely don't think we should go down the path of automatic translation.
RaisingK said:
I overlooked this topic and nuked
{ "无题", "無題", "Untitled", "rkgk", "練習", "Practice" }under the assumption that such common, generic strings didn't qualify as "interesting additional information" (and users might mistake them for such where untranslated). I shouldn't have acted unilaterally, though....would you prefer I add them all back?
I would prefer we keep them for statistical reasons, yes. Ideally we'd preserve all metadata - we're moving in that direction and in the future we'll also save things like pixiv/twitter tags, creation date from the source site, various exif etc.
A better way to handle them would be to mark commentary as irrelevant on site, rather than deleting them. I've opened issue #4459 for that.
