<?xml version="1.0" standalone="yes"?>
<Paper uid="T75-2001">
  <Title>AUGMENTED PHRASE STRUCTURE GRAMMARS</Title>
  <Section position="3" start_page="0" end_page="0" type="intro">
    <SectionTitle>
I. INTRODUCTION
</SectionTitle>
    <Paragraph position="0"> An augmented phrase structure grammar (APSG) consists of a collection of phrase structure rules which are augmented by arbitrary conditions and structure building actions. This basic idea is not new, having been used in syntax-directed compiling \[e.g.</Paragraph>
    <Paragraph position="1"> I\] as well as in natural language processing \[e.g. 2\], but what is new are the particular language in which these rules are written and the algorithms that apply them.</Paragraph>
    <Paragraph position="2"> This brief paper is intended to serve as an introduction to augmented phrase structure grammars. First, the form of data structure used is discussed, followed by discussions of the analysis and synthesis of text, i.e. decoding and encoding.</Paragraph>
    <Paragraph position="3"> (Although this session of the workshop is devoted to natural language inout, this brief discussion of synthesis is included because one of the important features of APSG is the consistent manner in which both decoding and encoding are specified.) Then there is a section on implementations and applications, followed by concluding remarks.</Paragraph>
    <Paragraph position="4"> II. DATA STRUCTURE The data structure used by APSG is a form of semantic network, consisting of &amp;quot;records&amp;quot; which are collections of attribute-value pairs. Records represent entities, either physical or abstract, such diverse things as vehicles, actions, words and verb phrases. There are three different kinds of attributes: relations, which have as their values pointers to other records; properties, which have as their values either numbers or character strings; and indicators, which have bit string values and usually serve in a role similar to features in linguistic terminology.</Paragraph>
    <Paragraph position="5"> A record that has a NAME attribute is called a &amp;quot;named record&amp;quot; and can be referred to by using the value of the NAME attribute in single quotes. Named records are used to hold information that is relatively permanent, such as information about relevant words and concepts, and are defined in the following manner (where the parentheses enclose structure-building information):</Paragraph>
  </Section>
class="xml-element"></Paper>