THE Web is pretty good at letting people surf through text, pictures, movies
and sounds. But if you鈥檙e interested in manipulating good, hard data, it can be
a bit of a let-down. Now new standards promise to allow people to go 鈥渄ata
mining鈥�, searching and comparing data from several sources.
鈥淲e want to make it simple to publish data so other people can do meaningful
things with it,鈥� says Robert Grossman, director of the Laboratory for Advanced
Computing at the University of Illinois in Chicago. Grossman led a group that
last week released the first version of the Dataspace Transfer Protocol
(DSTP).
The idea is to give people access to stores of data held on different servers
across the Net so that they can pull out the data they want from each one and
analyse it.
Advertisement
Suppose you suspected that sunspot activity might be related to global
warming. Searching the Web you might find a database of sunspot information and
another one of global warming information. But the chances are that the two
databases will be in different formats, and the data you want will be buried
under a mountain of material you don鈥檛 need.
Using the new protocol, you could transfer just the data you want from the
databases, align it by date, and call up a graph of sunspot activity against
temperature.
DSTP only works if data is presented in a new format called Predictive Model
Markup Language (PMML), just as the Web works using a method called Hypertext
Transfer Protocol (HTTP) to move documents written in Hypertext Markup Language
(HTML).
The new standards tell people who want to make their data available how to
format and label their columns and rows. Once the database is properly formatted
and labelled, a user can search and manipulate it with the DSTP.
The new, open standards are being developed and published by a coalition of
universities and businesses, including IBM, Oracle and NCR. They hope that, like
HTTP and HTML, they will be adopted voluntarily by a critical mass of computer
users.
鈥淚 think it鈥檚 what people for a long time have been looking for,鈥� says
Claudia Gardner of IBM鈥檚 Intelligent Miner for Data programme. Open standards
will help their applications become more popular, she says.
Grossman admits DSTP could encourage people to make dubious correlations,
such as plotting the incidence of heart attacks against the position of the
stars. 鈥淲hat鈥檚 to keep stupid people from doing stupid things? Nothing,鈥� he
says. 鈥淏ut this is to help smart people do smart things.鈥�