<?xml version="1.0" standalone="yes"?>
<Paper uid="M92-1031">
  <Title>CRL/NMSU and Brandeis : Description of the MucBruce System as Used for MUC-4</Title>
  <Section position="1" start_page="0" end_page="0" type="intro">
    <SectionTitle>
INTRODUCTION
</SectionTitle>
    <Paragraph position="0"> Through their involvement in the Tipster project the Computing Research Laboratory at New Mexic o State University and the Computer Science Department at Brandeis University are developing a method fo r identifying articles of interest and extracting and storing specific kinds of information from large volumes o f Japanese and English texts . We intend that the method be general and extensible . The techniques involve d are not explicitly tied to these two languages nor to a particular subject area . Development for Tipster has been going on since September, 1992.</Paragraph>
    <Paragraph position="1"> The system we have used for the MUC-4 tests has only implemented some of the features we pla n to include in our final Tipster system . It relies intensively on statistics and on context-free text marking to generate templates. Some more detailed parsing has been added for a limited lexicon, but lack of fulle r coverage places an inherent limit on its performance . Most of the information produced in our MUC template s is arrived at by probing the text which surrounds `significant' words for the template type being generated , in order to find appropriately tagged fillers for the template fields .</Paragraph>
  </Section>
class="xml-element"></Paper>