Feminist SF Wiki talk:Categorization

I propose Category:Folk & Popular Works for deletion because I created and relinked the relevant pages to Category:Folk and popular works. I've also added this proposal to an active admin's talk page in case no admins are watching this article. Contributor 18:18, 4 March 2007 (PST)

General categorization discussion

• picking up from discussion began on Talk:Race and feminist SF which briefly mentioned a "Category:Women of Color"

For myself (and until just lately User:Ide Cyan and I have been the main ones hanging out here for a while, so it's been a bit of an echo chamber for the two of us and our opinions), I've been puzzling over problems with category system and other ways of organizing, and what is the most useful way to use the mediawiki category system and other types of organizing. I started some categorization guidelines and discussion questions at FSFwiki:Categorization and FSFwiki talk:Categorization/Defaults to figure out how & when to use things as keywords versus hierarchical schema; how to handle questions of what are "defaults" and assumed (like on this wiki the default maybe should be women, but then is it weird to not have categories for "women writers"?)

... I think after a lot of experimentation I am leaning towards largely only using categories for either (a) hierarchical organization schemes that are obvious & discrete (like "year of birth") or (b) (still thinking through how this might work) limited numbers of discrete tags (like "Earthsea"). After following a lot of discussions on wikipedia, I've been figuring out that the category system works best when there are a fairly small number of categories on each article (<10 for sure). So if categories are used as tags, i *think* it will work best with a pretty discrete system of important keyword tags. These I have started putting at Category:Tags -- we could have many more, but I think a free-for-all in the classic tagging/folksonomy style won't work. (But even there have developed a bit of a subhierarchy, because set up "language tags" to tag things as written in whatever language.) The fact is, that the category system is not actually tags, so it can't generate weighted tag clouds or anything like that. It's sole built-in functionality is the creation of alphabetized lists and the ability to embed hierarchy. So, I think it's just limited in its usefulness. Think of it more as a controlled keyword list than a tagging system.

... In other words I'm thinking of categories as principally useful not for descriptive purposes but just for sorting ... I've been trying descriptive stuff, but I think it rapidly builds up into a big set of works, and the grey areas cause problems. For instance on Category:Works featuring female-only worlds, what about works in which men are introduced? or societies that are all-woman like amazons but there are men in the "world"? tags work fine for this sort of thing because if it's non-intuitive then most people won't apply it and the tagging software will incorporate that into the ranking. but categories are a binary, on or off -- there's no relevancy ranking or weighting. ... similarly, we can create Category:Writers for young adults, but if the writer is also a writer for adults and for children then boom we have three age-related audience categories, and that doesn't include any of the numerous other potential categories for that person (her three different ethnic heritages; her languages written in; her gender; her genres; her awards received; etc.). So you might think so what, let a million flowers bloom, but then it turns out that actually to use those categories from the author's page itself, they're useless when more than about 10 -- because they're not tags that weight by size or some other fashion if they are applied in many ways. ... And as an index of those individual features, the categories are less useful than lists because they can't explain nuances, or offer any way of sorting other than strictly alphabetical, etc.

For large swaths of intercategory stuff (books with X feature written by X kind of writer; X genre written by X writer; etc.) and for questions of tricky defaults (all female characters; all characters of color; no male writers; etc.) I think I'm coming to the conclusion that the category system isn't the best way to go -- it just lacks flexibility. So I'm using lists instead and also infoboxes -- these I think end up being much much more flexible. So a list of "Women SF writers of color" I think will be much more useful than the various categories that are starting to pop up (see Category:Writers by ethnicity, Category:Writers by nationality, etc. ... I just copied of List of female SF writers of color from the fsf.org website; it's woefully out of date & unlinked but it's a start. (Was your idea of Category:Women of color supposed to just be real women? also characters? what? I'd say, make a list of what you envision. redundant and overlapping lists are easy to deal with over time as we figure out what the scope should be.)

The other thing I've realized is that it is a PAIN IN THE ASS to unreconstruct categories if we decide that we went down a wrong direction. I had thought "oh why bother with wikipedia's capitalization rules" but then over time realized that i should have taken advantage of the years of experience with mediawiki software that wikipedians have accumulated and just fucking followed their example on that matter. once i figured it out there were over >1000 instances of one type of category and many, many more -- i'm still fixing them. so going slow on categories is GOOD because they seem so easy & trivial that people just start using them. they're easy to do, but a PITA to undo. whereas lists can be moved, split up, redone, retitled, edited, etc., very easily. ... and by using lists & infoboxes to start with, we can then ultimately figure out which kinds of labeling/tagging/sorting would be really essential and useful as categories. It's not about deprecating the value of the information embedded in potential categories/tags, but about realizing the limitations of the category system and figuring out other ways to embed the vital information. Categories work well for some kinds of information, and we have to figure out other ways (lists and infoboxes, and the search feature, what links here, and related changes, for now).

Hmm almost none of this was about race. which is fine, because these questions aren't really limited to any one label or identity but are important for figuring out how the technology works and is best suited for our labeling & identifying needs. --LQ 07:23, 25 April 2007 (PDT)

Hmmm okay - I know you are way more experienced in Mediawiki than I am. I tend to think in terms of tags. And the Category structure still confuses the hell out of me. I tend to want to make entries, tag them, and then construct lists from the resulting lists of pages that come up when I look for the stuff tagged with what I want to find.

For me it's also about how I search and how and where I make new pages. So I went to search for women of color etc. and could not find nearly anything and therefore started a list and a category, thinking that the category would develop out. Even after your explanation I don't understand why no category of women of color, but yes a category fo "african american writers" which I did end up finding? Would it be so bad to have that category? for example I would not mind having people tagged with multiple identifiers... So what if there ends up being a page of 300 "white writers" when you look at them as a category? Would that be so bad? I was trying to think in terms of being a person of color and just as I go to books and look up "women" I wanted there to be multiple pathways to finding that information.

I wish we had free tagging *and* your hierarchy in place simultaneously.

--Liz Henry 16:44, 25 April 2007 (PDT)

"African American writers" and "Writers" and "Women writers" and everything else in terms of the categories that exist have been created as experiments, and I keep tinkering with them to see what works. So that's why there's AAW, because it seemed intuitive that we needed and wanted it, because it's something we want to make available to people. Same with WOC. So it's not a question of what we want to make available, but how is the best way to make it available. I don't have a full philosophy of categories yet, but I'm starting to develop one. --LQ 18:54, 25 April 2007 (PDT)

Okay, more succinctly: "that the category system works best when there are a fairly small number of categories on each article (<10 for sure)."

Why?

--Liz Henry 16:45, 25 April 2007 (PDT)

I should have been more precise - categories work in several different ways. (1) On the article, you can look at the categories to see related categories; IMO it is visually very difficult to browse a list of categories if it's > 10. Look at the categories on wikipedia GWB. Maybe you disagree? For me, it's very hard to read through a list of unconnected phrases that are all hyperlinked. Just visually it's hard to stare at it and hard to make my eyes keep moving across the lines to read them all. So any one category can easily get lost there. But, if there are just a few then I can see them. ... The question then is, if just a few go on the article, how do you choose?

But there are other ways to use the categories, and maybe those would be useful enough to just say screw whatever utility one might find from looking at the list of categories on the article itself. These include: (2) Looking at the category itself, which is an automatically-generated alphabetical listing, and (3) Using the categories to navigate through hierarchies.

You mentioned #2, and I think it is helpful to have the automatically-generated alphabetical listing especially if it's as complete as possible. But that argues for a fairly tightly-controlled vocabulary of categories, so that you don't have, say, both "Black female writers" and "Female black writers", and some people in one, some in the other, some in both, and some in neither. Because for this to be useful it has to be complete. Folksonomy-style tagging gets around that because tags can cluster together or even as in librarything they can be related to each other or viewed as synonyms. But there's no way to do that with mediawiki categories.

So if you have a set of tightly-controlled vocabulary keyword categories, then you can be complete, and you have a great set of categories for browsing. There's a problem with big categories of >200 entries -- it's a user interface thing that does bug the shit out of me is that if there are >200 entries in a category, mediawiki goes to multiple pages; and it spreads the subcategories out evenly across all the pages. But it's not clear that subcategories are on multiple pages, so people think they're seeing all the categories on the first page; and that kills the hierarchical aspect of browsing. But that doesn't bother me too much.

The main problem that bothers me with categories, and I still don't have a good solution for it, is that it's really hard to figure out if items are missing from categories, and it's hard to see when items are added or deleted. More generally, the whole category system is just so damn inflexible. So, for instance, suppose I'm trying to go through and clean up the wiki and make sure that all the fantasy writers are categorized as fantasy writers. For example:

1. I can look at Category:Fantasy writers and see things that don't belong but it's much harder to see things that are absent, simply as a matter of interface because of how things list out--it's really hard to compare them. Lists can be formatted in all kinds of different ways that facilitate organizing (subgenres or last name or eras) and reading (extra spaces, subheadings, bolds; tables, columns, etc.) and sorting (you can sort different ways on lists, not always just alphabetical) and referencing (footnotes, explanations, subheadings).
2. If an editor adds a writer to Category:Fantasy writers I have no easy way of knowing that short of basically memorizing the list and recognizing changes to it. Same if they delete. (You can do "related changes" but I'm still not sure of the parameters of it and it doesn't seem to quite do what one would want or think.) Whereas, if it's a list, I can pull up List of fantasy writers and see the history and what exactly was added or deleted.
3. If there are pseudonyms or aliases or redirects from variant names, those don't show up either. Whereas on a list, you could include all the variant names and say "see james tiptree, jr." after "alice sheldon" and "raccoona sheldon" etc.
4. You can't fuck it up at all like you could with tags. Category:Women of Color is completely different than Category:Women of color. Articles you can redirect and they come up anyway with search because search is case-insensitive.

The hierarchy thing is a great add-on feature -- even if you don't want to use it, it's pretty cool to be able to list "writers" in a hierarchy of "people by occupation"; or "1905" in "20th century years" or whatever. But it's limited, I think, because it doesn't handle synonyms and redirects.

The "what links here" is usually kind of broken for categories, because most things don't "link" to categories. So using the category as the main way to tag something as "X" actually means that the tag is halfway broken, because you can't really see what links to it ... like, it would be better if there were just a paragraph called "tags" and each one was a hyperlinked keyword. then if you clicked on the hyperlinked keyword you could see "what links here" and it would be everything with that tag.

... etc.

there is something weird too about the categories that isn't the same with tags, either. because categories have this sorting function it can operate as a sort of de facto segregation. on wikipedia they wrestle all the time with people "diffusing" categories inappropriately. for instance "ethnicity occupation" (like "african american scientists") or "gender occupation" ("female scientist") are supposed to be duplicated, so that the person is filed still under scientists and african american scientists. but people don't get that, and they think oh great let's just eliminate redundant categories; since african american scientists is a subset of scientists we don't have to categorize that person under both. that whitewashes the main categories, which is clearly a problem, but it's hard to police it and see when it happens because categories are hard to police .... it's not necessarily that people are doing it for inimical reasons; they may just be enjoying editing the george washington carver article, and making a change they think is helpful only to that one article.

... so i'm experimenting with female characters and maybe ultimately female writers, creating a single giant category for all of them but making it part of an infobox so it isn't taken off by accident (Template:Femchar to see if this can get around some of the issues. but it wouldn't be a solution for all of the things we would want to categorize (like audience appeal, sexuality, ethnicity, nationality, language, politics, religion, ability, class identity, other important identities....).

having talked all this through (and sorry it's not concise at all) i now am thinking (a) maybe lists and infoboxes are still a good way to go for a lot of navigation, but (b) maybe we should also just build out the Category:Tags with things like "women of color". it would need to be vocabulary-controlled (because it's not a real tag system with relationships and ranking & proximity) but it might more closely approximate some tagging-like functionality. Just make a paragraph at the end or bottom of the page: ==Tags== [[Women of color]] [[Cyberpunk writers]] [[Blah3]] [[Blah4]] . If we needed to we could preface each one with "Tag" or "TAG" like "TAG cyberpunk writers" or whatever. then, even if they were different, we could do redirects; and "what links here" works with redirects. thoughts?

--LQ 18:54, 25 April 2007 (PDT)