Conclusion

Our experience with the MUC-5 evaluation leaves us believing more strongly than ever that the FASTUS system is the best approach to the text information extraction problem, and that the technology is now ripe for application to real-world problems. Our experience has shown that

• Information can be extracted from text very rapidly.
• It is possible to bring a system up to a high level of performance very quickly given the right set of tools. FASTUS provides us with the tools we need.
• The system architecture can be straightforward, conceptually simple and easily understood.

Future research in this area should be directed toward the application of this technology to increasing analyst productivity. Open questions are how a generic information extraction system like FASTUS can be customized to a particular application like JV-FASTUS without extensive intervention from its original developers. Interesting ideas to investigate are how a system can automatically acquire information from a corpus of related texts, or from learning and generalizing from observing analysts annotating texts.