|
|
About XTractor
Computers are powerful and intelligent tools
Information
is the currency of the modern world. Modern computers can
process and analyze millions or billions of individual
pieces of information every hour. They are the ultimate
power in processing information. They can find the
information you need, they can link new and existing
information, and they can alert you when something
unexpected is about to happen.
But they can't understand human language
Modern
computers can be powerful intelligent information analysis
tools, but only when the information is normalized and
encoded in a mathematically acceptable form: for example, in
a database record.
Unfortunately, this rules out the major source of
information that we humans use to communicate -- Free Text.
That's right, the all powerful computer is helpless
when is asked to extract information from an email, word
document, web page, etc. Why? Because the same information
can be represented in many different ways in plain text.
For example, how will the computer know whether
"White" is a person or a color? How will it guess
that "Infogistics" is a company and that
"Small Hopes" is a town? Indeed how?
XTractor
will solve this " communication
problem"
XTractor by
Infogistics is a successor to the
Named-Entity Recognition System developed
by our key scientists at the University of Edinburgh which
won the prestigious MUC-7 competition organized by the American
Defence Research Agency (DARPA) in 1998.
XTractor is an engine which sifts through large volumes
of texts and creates database records for the objects which
are mentioned in the text, such as people, organizations,
locations, vehicles, etc. Moreover, it normalizes and
unifies information, so for example, if somebody is
referred to as "John D. Smith", "John",
"J. Smith", or even "he", XTractor can
figure out that this is the same person. And most
importantly, XTractor identifies the key relationships or
links between the objects that it finds.
For example
For example,
from a sentence "White is a managing director of Cyber
Corp." it will identify that there is a person with
surname "White", there is a company called "Cyber
Corp." and it will create an "employment" link
between these two objects. It will also record particular
details of the employment: "managing director".
XTractor can be linked to a document Viewer where the
objects XTractor finds in are highlighted, and information
aggregated from many different places of a document can be
seen at a glance. For instance, at the beginning of a
document XTractor identifies that John Smith is 42 year
old, later on it appends that he works at Good Co. Ltd., at
the end of this document it identifies that he drives a red
Nisan. XTractor makes sure that all this information is
kept under the single record "John Smith",
suitable for further processing by the computer, or for
displaying to humans.
Interested? Want to know more?
Talk to us. contact
details
|
