Define a Robust Category System
May 1, 2006 10:33 AM   Subscribe

I'm developing an online "pop quiz" application as a hobby, and I just want to run the specifications of my projected categorization system by the askMeFi crowd for comment before I devote a whole lot of time to sorting questions into these categories, and writing questions to fit.

The Category Tree will be defined as follows:

(formatting didn't transfer, link to text file.)

With subcategories defined further when necessary or expedient, roughly according to the following example:

-->[Original Language]

Each question would reside in one subcategory, with automatic membership in parent categories.

A question specifically about Hamlet, for example, would be categorized as follows in a mature question set with tens or hundreds of thousands of total questions

Arts & Entertainment
-->-->English Literature
-->-->-->17th Century

The question, therefore, would be eligible for inclusion not only when the user selects the Hamlet subcategory, but also if the user selects any one of its parent categories.

My question is whether the categorization system as presented is broad enough to cover basically any English-langauge trivia question that could possibly be presented to the user.

Have I missed anything?
Does anybody have any other advice?
posted by The Confessor to Technology (4 answers total)
The categorisation isn't hierarchical, there could be things about shakespeare that aren't strictly about literature, and things about hamlet that aren't about the 17th century. Go with freeform tagging instead, and if necessary, organise those tags in a hierarchical directory.
posted by fvw at 10:39 AM on May 1, 2006

What about works that span eras? For instance a book like the Christian bible that's made up of sub-works -- I'm not 100% sure that the bible's books span centuries, but it probably does.

Then there is the more common case of works that have more than one author.

I suppose you could just allow questions to have more than one category.

Are you really asking whether the main categories given in your text file (I try to list it below) will cover everything?

I might suggest psychology, human relationships, food, fashion.

your list

Arts & Entertainment
Computer & Internet
Painting & Sculpture
Television & Film
Mathematics & the Sciences
Biology & Nature
History & Geography
posted by amtho at 10:56 AM on May 1, 2006

Have you considered a flat tagging system rather than a strict heirarchy?

You could have some required tags, say century-XX or country-XX, and let everything else sort itself out? This would allow you to create quizzes that contain West Side Story and R&J quite easily, which would make for more interesting quizzes and allow you to easily define better relationships.
posted by unixrat at 11:29 AM on May 1, 2006

Response by poster: amtho:

I apologize if my list seems terribly incomplete - being somewhat focused on the practical myself, I completely ignored such 'softer' subjects as Food and Fashion. I'll rectify that now.

As for multi-era books, since most such books are religious literature, I'm probably going to head such questions under a religion category - possibly under history.

fvw & unixrat:

The problem with a straight tagging system lies in my future plans for this application:

I'm hoping to come up with a difficulty rating system free of the Wikipedia/Google-effect by having contests in which the speed and accuracy of the person's response determine their overall score.

In such a scenario, in which adjoining questions could come from widely varying subjects, nested categories would seem to be a quicker, more certain method of establishing context than unordered tags...

Alternatively, context would need to be established textually in each question... which could become wildly repetitive for a person who chooses to specialize in a particularly narrow category.

Did that make sense?
posted by The Confessor at 12:03 PM on May 1, 2006

« Older Help me 2Rank my RAM choices   |   Sidewalk donuts Newer »
This thread is closed to new comments.