<?xml version="1.0" standalone="yes"?>
<Paper uid="C94-2212">
  <Title>NL Understanding with a Grammar of Constructions</Title>
  <Section position="7" start_page="1291" end_page="1291" type="concl">
    <SectionTitle>
5 Summary of results
</SectionTitle>
    <Paragraph position="0"> Our at)preach to NI,U is based both on linguistic arguments and on our dissatisfaction with the state of the art. State of the art systclns typically are too &amp;quot;syntax-driven&amp;quot;, failing to take context into account in determining the intended meaning of sentences. A related further weakness is that such sysreins are typically &amp;quot;sentcnce oriented&amp;quot;, rather than &amp;quot;conversation/diseourse oriented&amp;quot;, lit our view, this nmkes even the most robust systems &amp;quot;brittle&amp;quot; and ultimately impractical.</Paragraph>
    <Paragraph position="1"> To test whether a construction-based approach is feasible built a &amp;quot;complete&amp;quot; working system that would include a representation for constructions. To do this, we focused on the &amp;quot;calendar domain&amp;quot;, a domain with limited comi)lexity and simt)le but not nninteresting semantics. We have chosen to deal with simple actions, and not e.g. with question answering, where dee, per understanding would be necessary. 3 Our contributions: 1. We. have t)roposed a new kind of grammar computable construction grammars, which are neither semantic, nor syntactic. Instead, their &amp;quot;productions&amp;quot; combine lexieal, syntacti(', semantic and pragmatic information. 4 2. We have described data structures for constructions, and have shown that they can be effectively used by the parser. Note that the same data structure is used to encode tile lexicon anti the &amp;quot;syntactic&amp;quot; forms.</Paragraph>
    <Paragraph position="2"> 3. We have shown how to parse with~ constr~lctions.</Paragraph>
    <Paragraph position="3"> We have implemented a simple chart parsing algorithm, which carl be easily extended to an Eearlylike parser, as h)ng as tire construction granmlar remains L-attributed. We have found that even a simple parser of construction ear be quite ef\[icient. This is partly due to the fact that it does not require copying of all syntactic and semantic information from daughters to mothers; the goal of parsing consists in producing an interpretation, and strnetural information can be discarded once an intertn.etation of a phrase is produced. \]t is also worth e.mt)hasizing aWe have also thought about anoLher possibility, that is, enhancing an IR system, e..g. with the understanding of date expressions.</Paragraph>
    <Paragraph position="4"> 4In what sense are they &amp;quot;computable&amp;quot;? Althmtgh this adjective might sugge.st a formal model with computational complexity results, etc., what we have in mind ix pretty trivial: (1) the systcm actually computes the messages of grammatical construction; (2) the grammars and constructions are well defined data structures, and l)arsing (combining all associated constructions in all possible ways) is decidable.</Paragraph>
    <Paragraph position="5"> that invoking domain semantics drastically reduces the number of parses constructed.</Paragraph>
    <Paragraph position="6"> 4. We have proposed a modular architecture for NL interfaces based on the division between linguistic knowledge, domain knowledge base, and application knowledge base. Based on our experience, we believe that this architecture should work in general for speech-enabled interfaces for restricted domains.</Paragraph>
  </Section>
class="xml-element"></Paper>