<?xml version="1.0" standalone="yes"?>
<Paper uid="T75-2001">
  <Title>AUGMENTED PHRASE STRUCTURE GRAMMARS</Title>
  <Section position="8" start_page="0" end_page="0" type="concl">
    <SectionTitle>
VI. CONCLUDING REMARKS
</SectionTitle>
    <Paragraph position="0"> APSG clearly has much in common with other current computational linguistic theories, with the ideas of procedural specification and arbitrary conditions and strucutre-building actions being very popular at this time. It would seem to be most similar to Woods&amp;quot; augmented transition networks (ATN) \[5\], especially as used by Simmons \[6\]. Registers in the ATN model correspond closely to attributes of segment records in APSG, and the semantic network structures of Simmons are very close to the record structures of APSG.</Paragraph>
    <Paragraph position="2"> Context-free phrasestructure grammars have been known to be inadequate for describing natural languages for many years, and context-sensitive phrase structure grammars have not been found to be very useful, either. Augmented phrase structure grammars, however, appear to be able to express the facts of a natural language in a very concise and convenient manner, they have the power of computer programs, while maintaining the appearance of grammars.</Paragraph>
    <Paragraph position="3"> Although APSG was used successfully to implement one fairly large system (NLPQ), it is too early to do a thorough appraisal of its capabilities. Through the extensive use anticipated in the next year however, its strengths and weaknesses should become more apparent.</Paragraph>
  </Section>
class="xml-element"></Paper>