TextPro
TextPro is an information extraction system. It has been under development for approximately two and a half years, and has been extensively tested, and applied in a competitive software evaluation. It offers high performance in terms of speed and ease of domain-specific development, and is particularly well suited to information extraction tasks such as name-tagging.
TextPro has been used in several applications. It formed the basis of SRI International's name recognizer in the 1998 DARPA/NIST-sponsored Broadcast News spoken language system evaluation, (look for the results for the system sri-ie1, to see an example of its performance) and as the natural-language processing component of the SRI Maestro multimedia analysis and cataloging system.
TextPro got its start as classic guerrilla software. I developed
this program for the sheer fun of Mac hacking, and building something I thought
was both fun and useful. Its development received only very small amounts of
support from the government and SRI International. It didnt't take it very long
to evolve into a fairly substantial system with about 25,000 lines of code.
Its original purpose (aside from pure hacking pleasure) was to serve as a testbed
for the Common Pattern Specification Language, whose development was encouraged
by the TIPSTER Contractor Architecture Working Group. Many ideas from the TIPSTER
architecture have been adopted by TextPro.
Unfortunately, I am no longer able to offer TextPro as a free,
unrestricted download. I am currently working on a means to make the
system available to people in the research and educational
communities. Please bear with me while I work this out. Thank you for
your
understanding.
Douglas E. Appelt
October 10, 1999
for its contents. This is a safe-cache copy of the original web site.