<?xml version="1.0" standalone="yes"?>
<Paper uid="P92-1019">
  <Title>A CONNECTIONIST PARSER FOR STRUCTURE UNIFICATION GRAMMAR</Title>
  <Section position="3" start_page="0" end_page="144" type="intro">
    <SectionTitle>
INTRODUCTION
</SectionTitle>
    <Paragraph position="0"> The similarity between connectionist models of computation and neuron computation suggests that a study of syntactic parsing in a connectionist computational architecture could lead to significant insights into ways natural language can be parsed efficiently. Unfortunately, previous investigations into connectionist parsing (Cottrell, 1989, Fanty, 1985, Selman and Hirst, 1987) have not been very successful. They cannot parse arbitrarily long sentences and have inadequate grammar representations. However, the difficulties with connectionist parsing can be overcome by adopting a different connectionist model of computation, namely that proposed by Shastri and Ajjanagadde (1990). This connectionist computational architecture differs from others in that it directly manifests the symbolic interpretation of the information it stores and manipulates. It also shares the massive parallelism, evidential reasoning ability, and neurological plausibility of other connectionist architectures. Since virtually all characterizations of natural language syntax have relied heavily on symbolic representations, this architecture is ideally suited for the investigation of syntactic parsing.</Paragraph>
    <Paragraph position="1"> *This research was supported by ARO grant DAAL 03-89-C-0031, DARPA grant N00014-90-J1863, NSF grant IRI 90-16592, and Ben Franklin grant 91S.3078C-1.</Paragraph>
    <Paragraph position="2"> The computational architecture proposed by Shastri and Ajjanagadde (1990) provides a rather general purpose computing framework, but it does have significant limitations. A computing module can represent entities, store predications over those entities, and use pattern-action rules to manipulate this stored information. This form of representation is very expressive, and pattern-action rules are a general purpose way to do computation. However, this architecture has two limitations which pose difficult problems for parsing natural language. First, only a conjunction of predications can be stored. The architecture cannot represent arbitrary disjunction. This limitation implies that the parser's representation of syntactic structure must be able to leave unspecified the information which the input has not yet determined, rather than having a disjunction of more completely specified possibilities for completing the sentence. Second, the memory capacity of any module is bounded. The number of entities which can be stored is bounded by a small constant, and the number of predications per predicate is also bounded. These bounds pose problems for parsing because the syntactic structures which need to be recovered can be arbitrarily large. This problem can be solved by allowing the parser to output the syntactic structure incrementally, thus allowing the parser to forget the information which it has already output and which it no longer needs to complete the parse. This technique requires that the representation of syntactic structure be able to leave unspecified the information which has already been determined but which is no longer needed for the completion of the parse.</Paragraph>
    <Paragraph position="3"> Thus the limitations of the architecture mean that the parser's representation of syntactic structure must be able to leave unspecified both the information which the input has not yet determined and the information which is no longer needed.</Paragraph>
    <Paragraph position="4"> In order to comply with these requirements, the parser uses Structure Unification Grammar (Henderson, 1990) as its grammatical framework.</Paragraph>
    <Paragraph position="5"> SUG is a formalization of accumulating informa- null tion about the phrase structure of a sentence until a complete description of the sentence's phrase structure tree is constructed. Its extensive use of partial descriptions makes it ideally suited for dealing with the limitations of the architecture.</Paragraph>
    <Paragraph position="6"> This paper focuses on the parser's representation of phrase structure information and on the way the parser accumulates this information during a parse. Brief descriptions of the grammar formalism and the implementation in the connectionist architecture are also given. Except where otherwise noted, a simulation of the implementation has been written, and its grammar supports a small set of examples. A more extensive grammar is under development. SUG is clearly an adequate grammatical framework, due to its ability to straightforwardly simulate Feature Structure Based Tree Adjoining Grammar (Vijay-Shanker, 1987), as well as other formalisms (Henderson, 1990). Initial investigations suggest that the constraints imposed by the parser do not interfere with this linguistic adequacy, and more extensive empirical verification of this claim is in progress.</Paragraph>
    <Paragraph position="7"> The remainder of this paper will first give an overview of Structure Unification Grammar, then present the parser design, and finally a sketch of its implementation.</Paragraph>
  </Section>
class="xml-element"></Paper>