<?xml version="1.0" standalone="yes"?>
<Paper uid="A92-1024">
  <Title>Automatic Extraction of Facts from Press Releases to Generate News Stories</Title>
  <Section position="9" start_page="175" end_page="176" type="concl">
    <SectionTitle>
6. Conclusions
</SectionTitle>
    <Paragraph position="0"> JASPER shows that text understanding technology has progressed to the point that it can be applied profitably to real commercial applications. However, the state of the art will not allow the technology to be applied to unconstrained applications. Instead, applications must be selected carefully in order to yield positive results. Certain  characteristics of applications will make them better candidates for fact extraction using Carnegie Group's technology and, we believe, other technologies as well: * The events or reports have predictable components, or information types to be extracted.</Paragraph>
    <Paragraph position="1"> * The information to be extracted tends to be expressed through an unambiguous and predictable, though possibly wide-ranging, set of linguistic forms.</Paragraph>
    <Paragraph position="2"> Technologies like JASPER, which extract information from text using shallow, focussed processing techniques based on complex pattern matching and heuristic decisionmaking, can be profitably applied to applications having these characteristics. JASPER's accuracy compares favorably with other text understanding systems, and its processing speed allows for real-time use of the extracted information in a time-critical application. JASPER is a deployed system which solves a real business problem. We believe there any many other such business problems that could be solved with similar techniques. We expect to see the deployment of many more such applications in the future.</Paragraph>
  </Section>
class="xml-element"></Paper>