<?xml version="1.0" standalone="yes"?>
<Paper uid="W91-0223">
  <Title>The Autonomy of Shallow Lexical Knowledge</Title>
  <Section position="5" start_page="271" end_page="272" type="concl">
    <SectionTitle>
4 Conclusion
</SectionTitle>
    <Paragraph position="0"> A model which permits interaction between a syntactic module, a formal semantic module and world knowledge is theoretically attractive, and justified in psycholinguistic studies. World knowledge can be separated into a shallow linguistic layer and knowledge in  general. The linguistic layer contains just the information required for discourse interpretation. In the naive semantic approach to word meaning, words name concepts, and concepts are naive theories of objects and events. Experience building a computational text understanding system demonstrates that a constrained, predictable portion of these naive theories is sufficient to disambiguate words and structure, and to build a discourse model with anaphors resolved and coherence relations assigned.</Paragraph>
  </Section>
class="xml-element"></Paper>