Re: Inference search engines? -- links to inference searches included.
Roy Schestowitz wrote:
The problem with much of the content on the Web is that even dates are
hard-coded in unnatural ways. As oppose to XML, standardised metadata or
the like, dates are embedded as standard content, unlike LaTeX, for
instance.
There are some Perl modules that can help with this kind of data mining
until such time as there is a semantic web.
--'People have become' OR 'people are becoming' may infer
emerging developments from social points of view.
<LONG URL> >
--'Trend toward' AND 'becoming more' are one of the first
examples of keywords I developed for trend-watching.
Because of the very common practice of PR people to claim trends when
there aren't any, depending on the content of any few articles is not
likely to predict a trend. You would have to mine over hundreds of
articles, comparing content with your system. Since it is possible to
create temporary 'trends' with PR techniques, if you have enough money
to pump into it, careful weighing of sources is also in order.
Modification of weights given to particular sources might be one method
of training the system -- among others.
|
|
|
