A proposal about how to replace categories

[Larry Sanger]

Here is a definite proposal about how to replace categories.

"Finally," you say?  :-)

Background: I confess to a perhaps overextreme animus against Mediawiki's Category feature.  Yes, the feature is useful when it comes to categorizing articles for, say, workgroup assignments, or for categorizing images.  But when it comes to categorizing articles, well, the information contained on category pages is the same sort of information that one looks for in an encyclopedia article.  That is, a list of subtopics is perfectly valid as a sort of end matter as are lists of Further Reading and External Links.  This suggests, then, the following proposal.

PROPOSAL

(1) Delete all category information from CZ article pages except those links that assist workgroup management.  There might be some other exceptions; we will list them.

(2) If there was (or, if you would like there to be) a category titled C, then find the article titled C, and create an end-matter section titled "Subtopics".  You may also create "Related Topics" but only after "Subtopics."  This is because we want to encourage people by the structure of the end-matter to realize that they should not put subtopics under "Related Topics."

(3) (Totally optional.)  Coders are then strongly encouraged to build easily-navigable topic maps based on the links contained in the "Subtopics" and "Related Topics" sections.  Eventually, users might be able to opt to put useful strings of topic-subtopic relations somewhere on articles.  Suppose '>' refers to the topic-subtopic relation.  Then we say X > Y iff Y is in X's "Subtopics" section.  If articles in this way define a string of relations, such as A > B > C > D > E, then it's (theoretically) easy for coders to display those relations somewhere on the article E page--i.e., it's theoretically easy to display the "breadcrumbs."  And, of course, since there can be multiple parents for a single topic, there might be multiple "breadcrumbs."


ADVANTAGES

* (1) and (2) are dead simple to implement, while (3) is optional and can be deferred.  No new technical apparatus needed.

* We move information crucial to understanding a topic (namely, what other topic are related to it) to the page where the article page.

* We remove the redundancy that now exists between various lists and templates.


What do you all think?  What are the disadvantages?

I don't propose to make a final decision on this for at least a week, and I'd like to bring the community to as close to a consensus as possible.  Let's have a debate in the meantime.  Counter-proposals, one of which would be just to replicate the WP system, are welcome.

--Larry

[Zachary Pruckowski]

The immediate disadvantage is that this does not address lateral/horizontal relationships.  For instance, the relationship of Germany -> Berlin -> Reichstag can be covered by this, but the relationship of Berlin <--> Paris can not (at least as I understand it).

I would venture to say that horizontal relationships are more important to cover than vertical ones.  I would imagine most vertical links to be covered in the article text, while horizontal links are not as well covered.  For instance, an article about Impressionism will contain a Monet mention (WP's has it in the first paragraph or so), and an article about Monet will mention that he's an Impressionist painter.  The article on Impressionism could contain a list of prominent Impressionist painters (and it's easy, because there aren't a lot of them), but Monet's article will cover only a handful of his contemporaries.

I'm not saying we need all the horizontal links WP has (I rarely need to associate novelists by birth decade), but we have to make sure they're planned into the system.

[Larry Sanger]

The immediate disadvantage is that this does not address lateral/horizontal relationships.  For instance, the relationship of Germany -> Berlin -> Reichstag can be covered by this, but the relationship of Berlin <--> Paris can not (at least as I understand it).

Yes it can.  That's the business about "Related topics."

[Zachary Pruckowski]

The immediate disadvantage is that this does not address lateral/horizontal relationships.  For instance, the relationship of Germany -> Berlin -> Reichstag can be covered by this, but the relationship of Berlin <--> Paris can not (at least as I understand it).

Yes it can.  That's the business about "Related topics."

I'm not convinced that that can work as well.  "Related topics" branches a lot worse than "Sub-topics" or "Super-topics" could do.  What I mean is that in addition to having "Related topics" of "European capitals", we also might need "German Cities".  Now imagine an article like "Leonardo da Vinci".  There would be dozens of "Related topics" that relate to the main article, but are disparate from each other.

I'm not saying it's a bad idea, I'm just saying we need to sort out how to handle the large influx of data into the article itself.

[David Gerard]

Background: I confess to a perhaps overextreme animus against Mediawiki's Category feature.  Yes, the feature is useful when it comes to categorizing articles for, say, workgroup assignments, or for categorizing images.  But when it comes to categorizing articles, well, the information contained on category pages is the same sort of information that one looks for in an encyclopedia article.  That is, a list of subtopics is perfectly valid as a sort of end matter as are lists of Further Reading and External Links.  This suggests, then, the following proposal.

It's worth noting that several MediaWiki developers are working on turning the present MW categories system into something closer to "tagging" - with the ability to combine tags in a search.

[Shanya Almafeta]

Out of curiousity, wouldn't it be possible to do both this and categories simultaneously?  Perhaps it might be worth doing both, and let the readers decide how they're get to the next article they need.

Just a thought.

[Larry Sanger]

Out of curiousity, wouldn't it be possible to do both this and categories simultaneously?  Perhaps it might be worth doing both, and let the readers decide how they're get to the next article they need.

Just a thought.

Well, that's hard to support if we are committed to simplifying, not complexifying.

[Larry Sanger]

It's worth noting that several MediaWiki developers are working on turning the present MW categories system into something closer to "tagging" - with the ability to combine tags in a search.

Ouch--that would be a decided non-improvement.  The last thing WP needs is unprincipled, chaotic "tagging."

[Larry Sanger]

I'm not convinced that that can work as well.  "Related topics" branches a lot worse than "Sub-topics" or "Super-topics" could do.  What I mean is that in addition to having "Related topics" of "European capitals", we also might need "German Cities".  Now imagine an article like "Leonardo da Vinci".  There would be dozens of "Related topics" that relate to the main article, but are disparate from each other.

I'm not saying it's a bad idea, I'm just saying we need to sort out how to handle the large influx of data into the article itself.

There are all sorts of ways that good ideas can go wrong.  Obviously, in this case, we need to specify some rules about what goes in the "Related topics" section.

[tkjazzer]

I would enjoy more ways of viewing recent changes that are topic specific.  I'm not a fan of categories if they only list what articles are in those categories.  I want to be able to customize my recent changes to my preferences.  I wish there were some cool technology similar to some web services which track your preferences and then cater your search results.  I don't need my search results catered, only the recent changes options of seeing what I enjoy reading.  I like to see what is changing in CZ more than what was previously written.  I have WP to see what is already written until CZ grows more.

[David Goodman]

The problem is keeping track, and being consistent. For this to work well, experience in other information system has shown that it is necessary to maintain a list of headings and relationships, known as a thesaurus, and not rely completely of the people writing the articles.

Back at the beginning we discussed this from the viewpoint of the initial topics.  We got stuck at that point on what the main top-level topics should be, which is actually a relatively trivial point, and very easily changed. The hard part is keeping it all consistent.

People familiar with database structure will recognize these concepts. In information science, this is called an ontology. (WP heading is ontology (computer science) for it has a specialized meaning somewhat  different from ontology in philosophy). I think this is one of the better information science articles in WP, even if it does ignore work in allied fields like library science)

Now, the category system in WP seems to be improving, because there are people there interested in getting them consistent. In principle, this is no different than using the articles and linking between them, and the same problems arise. There is one key advantage of categories: which is the handling of very large groups. Let's take railway stations. In WP, you can go to the article,and see that there is a category, , and then go to subsets. With an article-based system, we would either have to creates a very long list at the main article, or actually write articles on "Railway stations in the UK" etc etc,  and then Railway stations in London, etc. before we got to the individual stations.

For a subject like biology there is no problem, because we will have articles on individual genera. But think of, say, US Congress Sessions, .

 

[Christo Muller]

I have not found categories much use in Wikipedia, relying rather on the Wikipedia&Google search in Firefox. (I have modified that for Citizendium: http://pilot.citizendium.org/wiki/User:Christo_Muller/Citizendium_and_Google_search_for_Firefox. Using categories, as far as writing an article is concerned, is for me like having to research the Dewey system when I wish simply to write a book. Internal links work. Related topics are not difficult to find with a decent search engine, and why should we then be doing a computer's work?

[Royalist0007]

There are many specialised subjects that require different forms of categorisation.The numismatic area is one of them.Here's an example of a typical article & its categorisation; http://pilot.citizendium.org/wiki/Banknotes_of_the_Chatham_Islands.

Aidan.

[Ori Redler]

The main problem with a category system is that if it is not perfect it is of very little use. It is also of very little use even when perfect.

The main reason for that, at least as I understand this, is that categories are tools to order items by proximity, not relevance.

For example, let's say that I'm interested in learning a little bit more about Feudalism. I open the category and find a collection of items related to Feudalism, but I have no useful clue as to what I should read first, what next, and so on. I can get a hint to this via |* (placing an item before the list) but that would apply only to one item. Dividing into sub-categories will not help here either.

That said, categories may still be useful if we regard them as nothing more than keywords. For example, if I look for "Beatles songs" I want to see Beatles songs, and if those are collected under one heading I may be aided in my search. Currently, the search engine is too lame to group the search results in any meaningful way that will help me here.

Or, in short -- categories may be useful, but only in a much-much-much-much flatter structure.


Ori

[tkjazzer]

what is use the most are the tables at the bottom of wikipedia articles.

scroll to the bottom of http://en.wikipedia.org/wiki/Sickle_cell to see the heme path articles.

http://en.wikipedia.org/wiki/Interleukin_1 scroll to bottom of this to see the different types of interleukins listed in table form.  As long as we keep these useful things.