Making IFDB's tag system more like VNDB?

I spent some time today thinking about tags on IFDB.

  • Today, anybody can tag any game with any tag.
  • Users can delete their own tags.
  • Moderators can delete anyone’s tags.
  • Moderators can also “block” tags, forbidding people from using a tag, in favor of a preferred synonym tag.

The Visual Novel Database (VNDB) is kinda like IFDB, for visual novels. Their tag system is quite different from IFDB’s tag system; VNDB is much more complex. https://vndb.org/d10

  • VNDB has a list of allowed tags. Anyone can file a request to create a new tag, but moderators must approve/deny the request. (It seems like they deny a lot of requests.)
  • VNDB organizes tags into a hierarchy. There are categories for theme, character, style, plot, and setting, and tags can contain “child tags,” e.g. “Dark Fantasy” is a child tag of “Fantasy.”
  • Users vote for and against tags on each game.

    Tags can be assigned a vote from 1 to 3. This vote should be interpreted as follows:

    • 1: The tag does apply to the visual novel, but is not too apparent or only plays a minor role.
    • 2: The tag certainly applies to the visual novel.
    • 3: The tag applies, is very apparent and plays a major role.
    • Additionally, if you believe that a tag does not apply at all you can also downvote it.
  • Tags with an average score of 2.0 or higher appear visible by default. Tags with an average score lower than that appear only on the “tags” tab.
  • You can also vote on how spoilery a tag is for a particular game. You can choose between:
    • “This is not a spoiler”
    • “This is a minor spoiler”
    • “This is a major spoiler”
    • You can also decline to vote on the spoiler status of a tag, leaving it “Neutral,” but the guidelines discourage using it.

    Some tags can act as spoilers for the plot of a VN, this can be indicated by voting on the spoiler status. It is highly recommended to not leave this as “Neutral”, and really try to give an indication of whether the tag is not a spoiler at all or does spoil a bit.

  • You can also vote on whether the tag is a “lie.” Lies appear as normal tags, but if you choose to reveal major spoilers, lies appear as strikethrough text, indicating that they’re not really true.
    • “This tag turns out to be false”
    • “This tag is not a lie”

    Tags are supposed to reflect the things you’ll be experiencing as you play the game, even if some things might get revealed to be lies or illusions towards the end of the story. In cases like this, add both the (arguably incorrect) tags and the tags describing the situation after the truth has been revealed, making sure to mark the “correct” tags as spoilers and the incorrect tags as lies.

  • VNDB also has a bunch of fancy user preferences.
    • There’s a “spoiler level” user preference.
      • No spoilers (This is the default)
      • Minor spoilers
      • Major spoilers
        (You can always click to reveal minor/major spoilers for any particular game.)
    • For particular tags, you can override its spoiler level in user preferences:
      • Keep spoiler level
      • Always show
      • Force minor spoiler
      • Force major spoiler
      • Always hide
    • You can also set an override display setting for particular tags:
      • Don’t highlight
      • Stand out
      • Grayed out
      • Custom color
  • VNDB also lets you define characters, and tag them with a special kind of tag called “Traits.” Each character can be linked to multiple games.
    • Traits aren’t voted on; users just edit Traits when they edit a character.
    • Traits do have a spoiler level and a “lie” checkbox.

I’m not sure I want to copy VNDB exactly, but these ideas are really interesting. I’ve filed a few issues on Github about them.

Questions I have:

  • Is anyone here in contact with VNDB admins/moderators? I’d love to talk to someone on the inside to get a sense for how they like it.
  • Are y’all familiar with other major wiki-style tagging systems? How do they work?
  • How do we feel about only allowing moderators to define new tags? Good idea, bad idea? (It might make it easier to keep tags organized. Would that be worth it?)
  • VNDB’s system is pretty complicated. Is that worth it?
4 Likes
What are you open to adopting from VNDB’s tagging system?
  • moderator-approved tags
  • hierarchal tags
  • tag applicability voting
  • spoiler rating for tags
  • lie rating for tags
  • user preference: spoilers
  • character tags
  • character trait tags
0 voters
What are you decidedly against adopting from VNDB’s tagging system?
  • moderator-approved tags
  • hierarchal tags
  • tag applicability voting
  • spoiler rating for tags
  • lie rating for tags
  • user preference: spoilers
  • character tags
  • character trait tags
0 voters

I’m most interested in dealing with spoilers, because that addresses an existing issue we have with tags.

Hierarchical tags could be interesting, but I’m not sure what difference it would make in practice. Would it just affect the arrangement on the page? Would people be able to hide one category but not another?

Re: moderator approved tags: That sounds like more work for the mods. Do they want to be in charge of that?

2 Likes

this seems really complex and idk how it would benefit ifdb significantly? VNs are way more popular than text IF, I’m not familiar with vndb but I wouldn’t be surprised if there were way more resources, mods, users, and usage there than here. what is the gain exactly and is it worth the overhaul and continued labor?

2 Likes

I don’t like the idea of making it so only moderators can create new tags; as mentioned, that would be a lot of work for the moderators, of which I don’t think IFDB has very many, and could become pretty frustrating for users.

I think the character tags/character trait tags are appropriate for a particular kind of game that you find more of on VNDB than on IFDB; if there were a database for specifically Choice of Games works, it might make sense there, but I feel like a large percentage of works on IFDB simply wouldn’t make much if any use of it.

In terms of voting on tag applicability, I’m more in favor of the idea you’ve brought up previously of having the tags be “owned” by the game and not the user and letting anyone delete a tag from a game, but I know there were a few people who were strongly against that, so if that’s no longer on the table, being able to vote on whether a tag is accurate enough would be a good alternative. However, I’m not sure it needs to be a multi-point scale as opposed to just an upvote/downvote thing.

I don’t know about VNDB, but how I’ve seen these tag hierarchies work is, if you search for a tag that has child tags, you will get everything that has the parent tags and everything that has the child tags. So for example, if “bird protagonist”, “canine protagonist”, and “feline protagonist” were made child tags of “animal protagonist”, doing a tag search for “animal protagonist” would pull up games tagged with any of those child tags even if they were not tagged with the parent tag. Which I think could be handy, although it’s not necessarily a huge priority.

3 Likes

FWIW, I was thinking about this in the context of making it possible for more folks to help remove “bad” tags without requiring moderators to do tons of work (or to entrust power in a lot of moderators).

There are a couple of filed issues about this:

#915 was going to be my preferred approach, but I’m kinda afraid of it, because it seems like an opportunity for edit wars.

In principle edit wars could happen for any piece of game metadata (author, description, external links, etc.) but people generally don’t go messing around with external links, because they are matters of objective fact. Tags are judgment calls. Is this game surreal? Is it horror or is it just spooky?

So, instead of just letting anyone delete a tag, my mind went towards letting people vote for tags (or against them). That’s quite nearly how tagging was originally designed on IFDB, and I was intrigued to see that VNDB has a sophisticated system for managing tag votes that we could copy wholesale.

A bit of history about how IFDB tag voting used to work

Originally, IFDB had this notion that multiple users could assign the same tag to a game, effectively “voting” for the game having that tag. So if a bunch of users all tagged a game “surreal,” then we’d know that the game was more surreal than games where just one or two users tagged it surreal.

IFDB even used to have a thing where it would show two numbers next to a tag, e.g. “IF Competition 2001 (3/22)”. There was a little help document that explained:

What do the tag numbers mean?

There are two numbers in parentheses listed after each tag in a game listing. For example:

IF Competition 2001 (3/22)

The first number is the number of other people who added the same tag to the game. (Why bother adding a tag someone else already added? It’s basically a vote for the tag: it strengthens the tag’s association with the game. This improves the search engine by telling it that the game is that much more relevant when someone searches for the tag.)

The second number is the number of other games with the same tag. You can see the full list by clicking on the tag.

But, quickly, it became clear that this isn’t how anybody actually uses tagging.

Once anyone tags a game “surreal,” there’s very little reason to tag it surreal again; it shows up in searches for tag:surreal if even one person tags it that way, and there’s never been a feature to sort by the number of people who assigned a game that tag.

Furthermore, this had odd effects:

Soon after tagging launch, MJR removed the two numbers, and just showed the second number, but left the schema in place, so users still own their own tags.


Separately there’s a question of consolidating tags. Tag consolidation is a never-ending Sisyphean task as long as users can randomly make up their own tags, that then need to be deleted and/or consolidated.

Rather than blocking known-bad tags, it might be easier to keep tags tidy if we switch to allowing only known-good tags.

Tag hierarchy is mostly useful to help with tricky consolidation cases. Should we consolidate “dark fantasy” with “fantasy”? They’re almost the same. Maybe all the games tagged “dark fantasy” should just be tagged “dark” and “fantasy”? Nesting “dark fantasy” within “fantasy” means that games tagged “dark fantasy” will show up in searches for “fantasy,” and then we don’t have to argue about consolidation any more.

2 Likes

Stack Exchange lets users vote to merge tags. On the sites I frequent there are never enough votes so mods just have to do it, but in theory the system could work. If the system was built for user voting but most of the time the IFDB mods just had to approve the suggestions then that would work if people don’t end up voting enough.

I’m not keen on the idea of “lie” tags. That seems way too complex, and surely most situations could be handled with other tags. For example, if a murder mystery actually turns out to be a faked death fraud, then it could just be tagged as murder-mystery with a spoiler tag of faked-death.

5 Likes

I really do not like lie tags, agreed

1 Like

If the IFDB had hierarchical tagging, do you really need to consolidate and prevent ‘bad’ tags in such a manner anyway?

Tag systems are never used perfectly. Being able to merge tags helps with synonyms, singular/plural tags, typos, abbreviations, US/UK spelling differences, etc. Especially if synonyms can remain after their initial merges, so that all future uses will also use the canonical tag.

2 Likes

In the case given by Dannii, I think that is covered by the “lying narrator” tag (sorry, I don’t remind the exact term used in that thread about gaslighting…) or I misunderstand the point ?

Best regards from Italy,
dott. Piergiorgio.

Well only if the narrator is lying from the start, which is not always the case.

Is “unreliable narrator” perhaps the term you were remembering from the thread? :slight_smile:

1 Like

yes, Luca, unreliable narrator. this was the term.

Best regards from Italy,
dott. Piergiorgio.

1 Like

I made a thread on VNDB’s forum to ask how people feel about the tagging system. It’s pretty interesting.

https://vndb.org/t22878

They seem pretty happy with a hierarchical tag list, managed by moderators, but it does require a lot of moderation work. The backlog for that work can take a long time (months, sometimes).

They’re also pretty happy with tag applicability voting, which prevents edit wars, but many games don’t have enough votes. (I think that problem would be even worse for IFDB.)

5 Likes

Someone in the linked thread had a good suggestion, and here’s my similar-but-kinda-different version of it:

An option that could encourage using preferred tags (and maybe encourage adding more tags in general) might be to offer a menu (I don’t mean a drop down menu–more like a form) showing the most useful standard tags, organized by category (genre, mood, etc.), where the user could just click on checkboxes or whatever to add each one. I’m imagining displaying a relatively small fraction of the tags that are in use on IFDB, since displaying a huge number of tags could be overwhelming, and since some popular tags (for example, “cover art”) are not that helpful in finding games.

Maybe there could still be an option at the end to type in new tags, but–again, similar to what the other poster said–if checking the boxes is low effort, it might cut down on new tags that are just very slightly different ways of saying the same thing.

(If we end up categorizing tags, maybe new tags could just be uncategorized by default.)

2 Likes

The VNDB thread is fascinating.

The thing that pops out for me (maybe this was obvious) is that tags have two major use cases.

  • Looking at a game page, trying to get a sense of what it’s like
  • Searching for games in some category

User-defined freeform tags are great for the first case and useless for the second case. (I am looking at post #6 by butterflygrrl https://vndb.org/t22878.6 in that thread.)

AO3 has both organized tags and “additional” (freeform) tags, because of exactly this. One story I’m fond of is about James Kirk and has “secret genius” as a tag, which is not something you’d go looking for but tells you something about that particular story.

4 Likes

The way IFDB works now, it’s possible for more than one person to add the same tag to the same game, but I’m guessing that’s not too common. If a prolific tagger closes their account at some point, are their tags lost?

If so, then maybe tag voting or making it easy to add duplicate tags (clicking options vs. typing them in) could have the benefit of making the system less dependent on any one user’s tags.

Yes, that would happen, but, to the best of my knowledge, it never has. If their account had existed for at least three months, it would be possible in principle for someone to use the IF Archive backups to generate a list of all the tags the closed account had made and resubmit them using the taggame API.