A different kind of knowledge mining.

Saco saco at moon.yerphi.am
Fri Mar 14 13:25:43 EST 2003


<gjgraham at concentric.net> wrote in message news:<b46rh1$guh at dispatch.concentric.net>...
> I have written a Perl program that reads scientific abstracts and
> joins fragmented assertions from them to make longer assertions.
>    Suppose that one publication says that Drug A increases Factor B,
> that a second publication says that Factor B increases Factor C, and
> that a third publication says that Factor C increases liver cancer.
> This would suggest that Drug A might indirectly cause liver cancer.
> The purpose of this Perl program (called Asserter) is to find and
> assemble such assertion chains, and bring them to the attention of
> the user.
>    Although Asserter is currently adapted to process scientific
> abstracts, in principle it could be adapted to process text from
> any source on any subject. I think that in the future, there will be
> many programs like Asserter in use to help people exploit the vast
> amounts of information available on the Internet. It amazes me that
> programs of this sort are not already available. (But maybe they are,
> and I have not found them.)
>    Asserter still has plenty wrong with it. It is slow, and chokes on
> large amounts of data. So, at this stage, it is only for the curious
> and the patient.
>    Asserter is available at www.xenobiotherapeutics.com. I eager to
> read peoples' comments and suggestions for improvement. 

Computer scientists have been working on similar problems for quate a
while[1]. Having such a tool would be great. If one can
interface it with prolog[2] it would be possible to make even more
complex assertions.

[1]http://citeseer.nj.nec.com/singh02open.html
[2]http://directory.google.com/Top/Computers/Programming/Languages/Prolog




More information about the Comp-bio mailing list