Making IFDB's tag system more like VNDB?

General IF Community Resources
21

I was curious for more detail and found this page. https://archiveofourown.org/faq/tags?language_id=en

Canonical tags appear in the filtering features on Archive of Our Own (AO3)—that means they appear in the autocomplete when creating works or bookmarks, and when searching and filtering for works. Tag wranglers give tags connections which put them into context. For example, we assign character tags to fandom tags and make them canonical so that when you post works for that fandom, characters belonging to it appear first in the autocomplete.

You aren’t required to use canonical tags for your fandoms, characters, relationships, or additional tags; however, using canonical tags ensures works and bookmarks show up in the correct filters right away.

See also the guidelines for AO3 “tag wranglers” https://archiveofourown.org/wrangling_guidelines

22

My current thinking is that I’d like us to incorporate some ideas from VNDB, but in a way that keeps moderator labor to an absolute minimum by allowing users to vote on way more stuff.

Here’s how I’m imagining it could work in my ideal world.

  1. Tag voting. We’d introduce score voting for tags, primarily to allow ordinary users to hide inapplicable tags without moderator intervention, but also to emphasize tags that characterize what the game is “truly about”, which could make it more feasible to recommend similar games.

    When you add a tag, you’d vote on its applicability, matching VNDB’s scheme.

    • 1: The tag does apply, but is not too apparent or only plays a minor role.
    • 2: The tag certainly applies.
    • 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.

    All existing tags would convert to a vote for level 2. Tags with a score of 2.5 or higher would appear emphasized (perhaps in a larger font); tags with a score below 2.0 would be collapsed by default. Users could click a button to “see more tags” to see them.

  2. Tag spoilers. We’d also have score voting for spoiler status. Spoilerific tags would be hidden by default, with a clickable button to reveal them, and a user setting to display them.

  3. Optional tag canonicalization. We’d have a list of “canonical tags” while allowing users to submit “additional freeform tags.”

    Following VNDB, canonical tags would have descriptions, parent/child tags, and any number of synonyms.

    Initially, a volunteer would canonicalize the top N tags. (Top 100? Top 1000?) Then, we’d allow ordinary users to charter elections on changes to tag metadata.

    • Add a new canonical tag
    • Add/remove synonym
    • Add/remove parent relationship

    All changes to tag metadata would be tracked in history, and would be reversible by a moderator or by a subsequent election.

    We’d design the rules of the tag-metadata elections to keep changes flowing quickly. e.g. we could say that each we’d close the election after 3 non-holiday weekdays if there are at least 3 upvotes and no downvotes, or after 10 non-holiday weekdays.

WDYT? Is this on the right track?

4 Likes
23

I haven’t spent much time on IFDB and haven’t used tags there at all, but I’m a big fan of the way Steam handles tags:

  • A game starts with whatever tags the developer chooses to put on it. All users can see these tags.
  • Anyone can tag any game with any free-form string
  • A user can always see and search by tags they’ve added themselves
  • A user-added tag doesn’t become visible to or searchable by other users until it has been added by some minimum number of users. Tags containing explicit language are not shown to other users

Based on the first post, it sounds like IFDB’s tagging system works mostly the same way, except for that last bullet.

The problem as I see it is that the more complicated the system, the fewer people will engage with it. This strikes a nice balance because the only tags you’re guaranteed to see are ones the developers/moderators and you yourself add, and then you’ll only see tags that some threshold of users agree on. This will likely remove spoiler tags as I imagine most users wouldn’t add those and the few that do get the benefit of seeing them, and there’s no opportunity for an edit war either.

This does leave open the possibility of many users adding a tag you don’t agree with, but Valve even addresses that in their Q&A on the feature:

Q. What if I don’t agree with a tag that has become popular for my game?
A. Tags can be a good indicator of when there is a mismatch between how you perceive your game, and how your game is perceived by customers. Often this is simply because there is some piece of information regarding the game that customers feel is missing from the store page.

Now what that threshold should be (50%? 20 users? Equal to number of users who have played it?) I have no idea.

24

And I suppose, for people who are particularly concerned, you could add a profile option to just never show user-added tags, only Developer/Moderator added and self-added ones.

Adding a whole exclusion system seems like a lot of extra complexity that I imagine would see limited engagement (says the guy who has literally hardly used the site at all :wink: )

25

What would the purpose of canonical tags be?

If we want to encourage certain tags–maybe make them easier to add, because we think they are going to be very useful–we might not just want to choose the top N tags. We might want to approach it in a more organized way.

In my opinion, some of the most common tags are not the most useful ones. For example, “cover art.” It might be that some tags are common because they are easy to apply–you can tell if a game has cover art, or if it was in a certain comp, without ever playing it.

How would the synonyms be used? When we added the feature for moderators to bulk add tags, I thought there was a concern that it would be more expensive to do searches when there are synonyms to take into account, or something like that.

1 Like
26

At IFDB, we currently have a pretty flimsy tagging system. Anybody can make up tags, and tag any game with anything. All tags are visible to all users, and searchable. Users can delete their own tags, but only moderators can delete other people’s tags.

I did think about Steam tagging; I feel that Steam gets a ton of mileage out of the fact that there’s a developer who blesses an official tag list.

For IFDB, developers don’t normally create/edit their game listings; volunteers do. And we don’t have a critical mass of users tagging games; we’re lucky if even one user adds any tags to a game. And we don’t have enough moderator time available for moderators to do all the tagging.

Because our traffic is so low, if we said “at least three users would have to tag a game as Horror for it to show up with the Horror tag,” hardly any games would have the Horror tag, even games in horror-themed competitions like ECTOCOMP.

We don’t even have a way to verify that a given user is the author of a particular game. Right now, that’s fine, because if you claim that you wrote a game, all it would do is link to your user profile, but if it gave you special tagging powers, we’d need to be sure that you’re really the author before we let you do it.

1 Like
27

Canonicalization is a necessary prerequisite for parent/child hierarchy, categorizing tags, and synonyms.

I generally call those things “tag metadata,” and if we want any of it, we can’t just have users type freeform text in a box. The tag needs to have row in the database where the category, relationships, and synonyms get decided. And the tag needs to have a chosen, primary name, the name that we would suggest in autocomplete.

As for the benefits of hierarchy/categorization, it helps a lot with the ability to browse the tags. Tag index | vndb It also helps that you can search for “fantasy” and get games tagged “dark fantasy.” Plus, I think knowing that tags are related will be an enormous help for a “game similarity” system based on tags.

On the VNDB thread I linked, the main dev Yorhel wrote this about hierarchy:

Organizing tags in a hierarchy was definitely a good decision. It greatly helps with discovery and provides an intuitive “this tag implies that tag” system, which in turn helps using tags for searching: users can find Comedy VNs even if it has only been tagged with Parody; no need to manually add such duplicate tags to the VN.

1 Like
28

If the traffic is that low, two alternative thoughts come to mind:

  1. I wonder if it’s worth changing anything to begin with. A more robust system feels nice because it has structure and process, but is there evidence that a significant amount of IFDB’s traffic is experiencing issues due to the limitations of the current system? If not, why add a bunch of complexity to solve issues people may not be having?

  2. Would it make sense to change the ability of users to add games, in favor of a game request form that moderators would approve? It means mods would need to review games before they get added, but they could add whatever basic tags at the time. No idea what the ratio of new game additions to new tag additions is and which would be more of a hassle for mods.

29

Let me try asking this a different way: What is the intended difference between canonical tags and non-canonical/freeform tags? Is the idea that most non-canonical/freeform tags will eventually be made canonical? Or would we intentionally have only a smaller group of tags be canonical because we’ve decided they are the most useful, standardized, or whatever, and we make them canonical because we want to encourage them to be used?

1 Like
30

That seems like a system that would make it particularly hard to reach consensus on how to spell tags for certain concepts, what to tag, etc.

(In my experience, developer-added tags also often tend not to be sensitive to existing conventions. See also: the Wild West that is the “tags” field in my ebook library, outside of the specific tags I add myself.)

31

Canonical tags are tags that are organized and standardized. These are for people to find games with and classify games. Examples: science fiction, source code available, parser, sexual content, walkthrough.

Freeform tags are more game-specific. They can be used for comedy, or just for more specificity that is too niche for a canonical tag. Most have only one or a few uses. Some of these can be eventually canonical, but some are not meant to be. Examples: big crunch, beetrootphobia, Castilian, Captain Nemo, the world has begun to warm. The mosquitos will arrive soon.

Does that clarify things?

32

I think there’s a general consensus that people would like to see the IFDB tag system cleaned up and made to function better—see past discussion here, for example: IFDB TagFest 2024 (information and guidelines)

4 Likes
33

That seems like a system that would make it particularly hard to reach consensus on how to spell tags

The idea is that it doesn’t matter. You don’t need a consensus because the user threshold makes it a de-facto consensus. If enough users add the same tag with the same spelling, it becomes available to others. If enough users then add an alternate spelling tag, then that becomes available too. The behavior of the users dictated that both spellings are of interest to people… or not.

34

I’d prefer to have an actual duplicates system, and then just have the IFDB committee settle on a spelling system (probably US, even though it’s not what I’d use myself.)

1 Like
35

Are you saying this is how VNDB does it?

I was trying to figure out what @dfabulich saw as the intended use of the canonical tags vs. the intended use of non-canonical tags when both kinds of tags coexist on IFDB. That might or might not be the same as the way they are used on VNDB–I don’t know.

36

Here are the top 100 tags on IFDB, listed loosely and imperfectly in categories. I was trying to see what kinds of categories were represented among the top tags. (Maybe I should have started a separate thread?)

Genres

adaptation (130)
Adventure (157)
comedy (187)
fantasy (866)
historical (157)
horror (776)
humor (473)
Mystery (418)
romance (286)
rpg (470)
science fiction (685)
slice of life (183)
Surreal (305)
visual novel (136)

Authoring systems

adrift (485)
Adrift 4 (303)
adventuron (160)
ChoiceScript (182)
ChooseYourStory (232)
Inform (174)
Inform 6 (135)
inform 7 (313)
TADS (491)
tads 2 (366)
twine (1338)
Eamon (280)

Competitions

Ectocomp game (410)
IF Short Games Showcase (163)
IFComp Game (1418)
IntroComp (173)
La Petite Mort (164)
Le Grand Guignol (161)
Mini-comp (186)
neo-twiny jam (256)
Neo-Twiny Jam 2024 (137)
Spring Thing Game (296)
speed IF (139)
speedif (330)

Protagonists

child protagonist (186)
female protagonist (748)
gender-neutral protagonist (454)
male protagonist (583)
multiple protagonists (205)
nonhuman protagonist (408)
teenage protagonist (167)

Content/subject matter

dragon (160)
LGBTQ+ (255)
magic (251)
profanity (245)
sexual content (283)
space (148)
violence (459)
Xyzzy response (535)

Parser/choice/interface

CYOA (797)
choice-based (766)
hypertext (436)
parser (1764)
primitive parser (128)
two-word parser (715)

Prose characteristics

first person (345)
present tense (188)
second person (341)
third person (175)
word count limit (313)

Technical features

bookmarks (197)
built-in hints (351)
gender choice (188)
landscape mode (196)
score (231)
undo (152)

External Resources

ClubFloyd transcript (680)
I6 source available (224)
I7 source available (197)
source available (753)
walkthrough (1877)

Types of Gameplay

combat (144)
conversation (132)
maze (195)
puzzleless (198)

Multimedia

cover art (1689)
graphics (1131)
music (342)
sound (621)

Language

foreign language (268)
French (170)
german (216)
Spanish (173)

Platform

C64 (181)
desktop (131)
zx spectrum (145)

Author information

collaboration (148)
first effort (194)

Award/Prize

Colossal cash prize (173)
Winner of a comp (313)

Game/Story Structure

linear (151)
multiple endings (702)

Setting

single room (253)

Length

short (810)

License

commercial (505)

Publisher

choice of games (165)

1 Like