<?xml version="1.0" standalone="yes"?>
<Paper uid="J80-2003">
  <Title>Responding Intelligently to Unparsable</Title>
  <Section position="10" start_page="0" end_page="0" type="relat">
    <SectionTitle>
5. Related Work
</SectionTitle>
    <Paragraph position="0"> Several other projects have concentrated on giving meaningful responses to partially understood input and of correcting erroneous assumptions.</Paragraph>
    <Paragraph position="1"> Kaplan (1977,1978,1979) reports on research which extends the notion of presupposition. Furthermore, he has developed algorithms for computing the extended notion called presumption, particularly taking advantage of the simplifying aspects of natural language queries of a data base. The algorithms give helpful responses to data base users when the query as stated would have the empty set as a response. Mays (1980) deals with presumptions related to users' perceptions of the logical structure of a data base.</Paragraph>
    <Paragraph position="2"> Codd, et.al. (1978) describes the first version of a system called RENDEZVOUS, specifically addressing the same problems as our paper, but proposing very different approaches. Unlike the ideas presented here, RENDEZVOUS is aimed only at interfaces to relational data bases. It provides many interesting human engineering features for clarification dialogue, even to a menu-driven specification of a formal query when natural language queries prove unsatisfactory.</Paragraph>
    <Paragraph position="3"> Some very promising work which is complementary to ours is reported in Hendrix, et.al. (1978) and Hendrix (1977). They report on a new software tool LI-FER, which enables rapid development of semantic grammars (Burton, 1976 and Brown and Burton, 1975). LIFER provides some error messages for unparsable forms by printing the possible items that could appear at the point where the parser could not proceed. Their heuristic for selecting one place where the block occurred is similar to ours. Combining the following additional features of LIFER with our work could offer a powerful natural language interface. LIFER allows naive users the ability to add synonyms for previously known words and to define new syntactic forms for sentences by the user presenting a sentence in the new form and an equivalent sentence which is already parsable. It also provides an automatic facility for handling ellipsis.</Paragraph>
    <Paragraph position="4"> Kwasny and Sondheimer (1979) have extended our notion of selectively relaxing predicates to deal with American Journal of Computational Linguistics, Volume 6, Number 2, April-June 1980 107 Ralph M. Weischedel and John E. Black Responding Intelligently to Unparsabla Inputs co-occurrence violations and relaxation of expected word categories. Their paper also reports a uniform way of treating ellipsis and conjunction, including gapping.</Paragraph>
    <Paragraph position="5"> Allen (1979) argues that good clarification dialogue requires that the system have a model of the plan the user is following and of how the sequence of speech acts by the user fits into that plan. We agree, and one of our long-term goals is use of a model of user goals, plans, and speech acts for this purpose. Other computational models of speech acts appear in Cohen and Perrault (1979), Levin and Moore (1978), and Mann (1979).</Paragraph>
    <Paragraph position="6"> Pattern-matching as an alternative to a top-down, left-to-right parser, has often been suggested as a means of processing ill-formed input, as discussed in Kwasny and Sondheimer (1979), for example. Hayes and Reddy (1979) also advocate pattern-matching as a part of an approach that they are implementing to cover the broad spectrum of problems in graceful interaction, including anaphora resolution, explanation facilities, flexible parsing, generating descriptions of entities in context, monitoring focus, and robust communication. null</Paragraph>
  </Section>
class="xml-element"></Paper>