Danbooru

commentary_request tag is being misused

Posted under Tags

punished K

Continued from post #5879420:

Not sure who else is running them, but bots adding the artist commentary + commentary_request tag are not a good service.

e.g. My upload post #5863425 already had english_commentary, then a bot adds the incorrect commentary_request tag.
From the wiki:

The artist's commentary from the original site has been posted, but it has not been translated into English.
(emphasis mine)

I'm not going on the Discord to discuss this.

(tagging @redtails )

Reply
Unbreakable

Mexiguy said:

I'm gonna play devil's advocate and say that at least in that particular post case it was missing the commentary tag (which according to the wiki should be used even if the commentary is in english to begin with), so it probably got caught in a -commentary search.

Yeah, 99% sure the bot looks if the post is tagged commentary request or commentary and adds the former if neither is there.

Also, while I don't support mass-adding commentaries like redtails does, I think it's worse to do it without adding any of the commentary tags.

Reply
punished K

Unbreakable said:

Yeah, 99% sure the bot looks if the post is tagged commentary request or commentary and adds the former if neither is there.

Also, while I don't support mass-adding commentaries like redtails does, I think it's worse to do it without adding any of the commentary tags.

Ah, that would make sense.

In that case, I need to add a BUR to this thread.

Reply
redtails

These are the possibilities

1. No bot, no auto-import of artcomm, many artcomm will be missed and eventually unobtainable due to bad_id
2. Bot, but no comentary_request
3. Bot, but no language processing, thus auto-tagging commentary_request
4. Bot with language processing

Let's go over each option, my comments.
1. Missed opportunity. Danbooru is concerned with art and its metadata. Leaving out metadata is ultimately bad for our cause.
2. and 3. This metatag really shouldn't even exist, we already have commentary:untranslated. People asked me to add commentary_reqeust because omitting it is wrong, so I did.
4. Perfect solution, but requires extensive testing and its error rate may in fact be similar to the existing error rate i.e. adding coommentary_request to a post with english_commentary or symbol-only_commentary

We've gone over this a number of times already with no solution. If you have anything to add beyond what is already said, go ahead. Otherwise, I suggest this topic is closed. My recommendation to you, as it seems to me that you are concerned about the use of that tag, is to open a BUR to request to nuke the commentary_request tag. We already have commentary:untranslated so in practice we don't need a separate metatag for it.

Reply
Super Affection

I vote for option 2: Run the bot so we don't lose anything, but don't add commentary_request automatically.
I may be in the minority, but I see commentary request differently from "this post has words that aren't in English attached to it". Reasoning:

  • Many commentaries are short and meaningless. Lone character/copyright names, various words for "sketch", "collection", and so on. I wouldn't tag those with commentary even if translated, because they'd only add noise to the tag. Tagging them as commentary request when untranslated is similarly harmful. I don't have hard data, but from my experience I wouldn't be surprised if most commentaries fall in the "uninteresting" category. I'm well aware that "short" doesn't imply "meaningless" and I wouldn't argue with someone adding commentary request to a post with a one-word description they don't understand because it might be meaningful, but doing so automatically would have an unacceptably high amount of false positives in my opinion. In other words, commentary request should only apply where commentary would (or might) apply once translated, and that isn't nearly true of all untranslated commentaries.
  • Translation manpower is limited and it isn't realistic to expect that all non-English text on the site will get translated. The request tags are more useful if they are only applied where the text in question has a good potential to be meaningful and/or someone has a particular interest in reading it, so translation effort can be focused where it's most helpful.
  • As already mentioned, the commentary:x metatags work for the technical interpretations of these tags, and are more reliable if that's what you're looking for. If the manually maintained ones don't have any separate meaning, why have them at all? Adding commentary_request unconditionally with a bot makes it redundant with commentary:untranslated. The tag only has meaning if it's added by hand.

Separately, it's worth pointing out that auto-tagging symbol-only commentary does not require language processing. A simple Unicode category lookup would suffice. If every character is either some type of symbol or whitespace, and at least one non-whitespace character is present, that's symbol-only commentary and will catch 99% of the tag's application, only missing stuff like kaomoji and not having any false positives.

Reply
nonamethanks

Super_Affection said:

Many commentaries are short and meaningless. Lone character/copyright names, various words for "sketch", "collection", and so on. I wouldn't tag those with commentary even if translated, because they'd only add noise to the tag. Tagging them as commentary request when untranslated is similarly harmful. I don't have hard data, but from my experience I wouldn't be surprised if most commentaries fall in the "uninteresting" category. I'm well aware that "short" doesn't imply "meaningless" and I wouldn't argue with someone adding commentary request to a post with a one-word description they don't understand because it might be meaningful, but doing so automatically would have an unacceptably high amount of false positives in my opinion. In other words, commentary request should only apply where commentary would (or might) apply once translated, and that isn't nearly true of all untranslated commentaries.

FYI symbol-only commentary implies commentary, so it's kind of a moot point.

Reply
Unbreakable

Super_Affection said:

  • As already mentioned, the commentary:x metatags work for the technical interpretations of these tags, and are more reliable if that's what you're looking for. If the manually maintained ones don't have any separate meaning, why have them at all? Adding commentary_request unconditionally with a bot makes it redundant with commentary:untranslated. The tag only has meaning if it's added by hand.

The one thing the metatags can't do is match the partial commentary tag, otherwise they work well enough yeah but the commentary tags should be use properly while we still have them around, commentary request for any untranslated commentary and commentary for fully translated ones.

Reply
LQ

Bots should never add commentary request. Contrary to its wiki, you should only add it when you want someone to translate the commentary.
(Most of the time you kind of do kind of want to know, and the site would be better if everything was in English, but...)
In practice, it's best to not waste "request space" when you don't actually care all that much.

+1 for nuking commentary request as commentary:untranslated exists; and if it's not adequate, it should be changed to be adequate. Ultimately, translators will just translate whatever commentary they want to, as commentary request is kind of meaningless. It's just more "user friendly", being a pretty yellow tag that even new users see all the time. Would be neat if after a nuke the autocomplete would resolve commentary_request to commentary:untranslated as a sort of alias.

Over time, anyone that actually actively/seriously translates random untranslated commentary will figure out commentary:untranslated eventually, and hobbyist/occasional translators will just translate posts that they (a) upload, or (b) discover have no commentary, as they do already.

Updated

Reply
user 499293

I agree with LQ, I don't really like having things like this be meta tags when there could be a better solution. It's not an actual request tag. Ideally it would just be a field on the commentaries table indicating whether or not it's translated. We could also have a field for language and do away with the language_commentary tags.

Reply
1 2