<?xml version="1.0" standalone="yes"?>
<Paper uid="H89-2006">
  <Title>UNIFICATION-BASED SEMANTIC INTERPRETATION IN THE BBN SPOKEN LANGUAGE SYSTEM</Title>
  <Section position="3" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 INTRODUCTION
</SectionTitle>
    <Paragraph position="0"> Over the past year work on semantic interpretation in the BBN Spoken Language System has shifted from a Montague Grammar (Montague, 1973) style rule-for-rule approach to one which attempts to carry out semantic interpretation directly in the unification grammar rules themselves. This is accomplished by adding semantic features to the grammar rules, placing them on the same footing as the existing syntactic features. Meaning representations are thereby constructed, and semantic filtering constraints applied, as part of parsing the utterance.</Paragraph>
    <Paragraph position="1"> We view such a move as having essentially three advantages: null * more information is available to semantic interpretation, so it is possible to gain higher coverage * syntax and semantics are integrated, so semantic filtering constraints can be applied as constituents are built and attached * this integration is simple and does not require any complex engineering of cooperating software modules null All three of these advantages are important ones for a spoken language system.</Paragraph>
    <Paragraph position="2"> The HARC system has the following overall organization. Spoken input is initially analyzed by the &amp;quot;N-best&amp;quot; algorithm(Chow and Schwartz, 1989), converting it into&amp;quot; a rank-ordered set of N best word-sequence hypotheses (for a given value of N). These N hypotheses are then analyzed by the parser, using the combined syntactic/semantic grammar. Those hypotheses which are syntactically and semantically allowed emerge from the parser as initial logical forms in which quantifiers are interpreted &amp;quot;in place&amp;quot;. Next, the quantifier module assigns scopes and passes the translation to the anaphora component, which then resolves the referent of intraand extra-sentential pronouns. The completed logical form is then passed to the back-end component whose responsibility is to compute the appropriate response to the user's input.</Paragraph>
    <Paragraph position="3"> The present paper confines itself to a description of the combined syntactic/semantic grammar, along with some discussion of the parsing algorithm. We will first consider the representational framework in which the grammar is written.</Paragraph>
  </Section>
class="xml-element"></Paper>