I am thinking of components of a new kind of database tool. It is part and parcel of how I envision a multitouch interface becoming a new paradigm for computing. The initial question is "How do I deal with millions of things?". Then considering a 'glasstop' multitouch GUI what kind of primative and second-order operations might gestures generate.
One class of these deal with enumeration, and I'm thinking of this in the context of data mining. The first thing I want to do is generate some human-comprehensible names for result sets, one dimension at a time. My paradigm is naming conventions. So, imagine that we have a large set of data and upon this data we run a random transform - what do we name the set?
If we go back to the deck of cards data and the 'hoop' operation that I mentioned in 'glasstop', we recall that I may have 1 million 'records' and that when I set them through this transform I get, say, 15 new sets of records. I have distributed them and now the 15 new smaller decks will divide the larger set. The distribution should be named and tracked, so now we need 15 names for the smaller decks. Each of these 15 will can be reduced to a histogram of the gating/distributing function.
The Enumerator is a labeling / indexing function that assigns familiar human comprehensible attributes to result sets. The cardinality of the distribution is matched, along with user preferences to a predefined or dynamically generated set.
Consider a deck of cards that goes through a hoop called 'birthday month'. The deck of cards represents a demographic database with 1 million persons. Of those million, 89% have a birthday attribute. The function will then distribute them to 13 piles with the 13th being "N/A". But we could assign an enumeration target not to 'birthday month' per se, but to a color, or a color and shape combination or to an animal. 'January' could be mapped to 'Blue' or any such set with a cardinality of 13. This gives the human operator a way of cognizing and associating intermediary datasets at an abstract level. Alternatively the transforms can be transparent with those functions and their result sets directly named without abstraction. The abstract enumerations can be turned on and off at any time.
Clearly, size, shape and color are the most primary and recognizable abstract enumerations.
--
Comments