I have been involved in a number of discussions concerning
data
and text mining recently and wonder if anyone has any
experience
with these topics that they would like to share. The basic
question is whether the license for an electronic resource
in a
form suitable to be read by humans extends as well to a
license
for machine-reading.
The area of data and text mining for scholarly materials is
a new
one, at least to me. My understanding is that materials
(research data, user data, published articles, books, etc.)
can
be gathered together in such a way as to enable robots to
sift
through them and identify patterns and themes. These new
patterns--effectively robot-generated discoveries--may
include
things that are not present in any single document in the
collection. Thus, the collection is greater than the sum of
its
parts, but that greater value is only perceptible by
machines.
This past week I heard an excellent presentation (it is not
yet
online, but when the link becomes available, I will post it)
by a
biostatistician, who commented that human access to such
databases is "of low value," in contrast to the
"higher value of
robot access."
Data and text mining are sometimes being discussed in the
context
of the idea of "Web 2.0," but I think this is a
mistake. Web 2.0
is a concept of Tim O'Reilly's to describe the emerging
practices
on the Internet today in the areas of community-building and
user-generated content. Web 2.0 is a metaphor, not a
technical
specification--but a very valuable metaphor. O'Reilly, for
example, distinguishes between the early Web (his 1.0) and
the
evolving Web by contrasting Encyclopaedia Britannica and the
Wikipedia. Both 1.0 and 2.0, however, share the fact that
the
users are humans. Data mining is a game for machines. It
would
be inaccurate to call it "Web 3.0" because
machines don't require
a Web interface at all. Web 2.0 is post-modern, but
data-mining
is post-human. Today's neologism: the Post Human
Internet, or
PHUNET for short, pronounced either FOO-net or (my
preference)
PEE-YOU-net. See Charles Stross's novel Accelerando.
Whether or not database mining of this kind will yield the
kind
of new insights some believe it will, I do not know, but it
would
be useful for the rights situation to be clarified early on
to
fend off litigation at a later time. It seems likely to me
that
publishers will begin to separate human- and
machine-readable
rights, just as they distinguish between subscriptions for
libraries and individuals. There is an interesting
precedent put
forward by some members of the library community, who argue
that
it is reasonable for publishers to charge for hardcopy, but
electronic materials should be free. It is conceivable that
over
time the "low value" of human-readable rights
will become Open
Access, leaving the higher value PHUNET rights for
aggressive
economic exploitation. It boggles the mind to think what a
large
collection of science articles could be worth some day.
Joe Esposito
|