<?xml version="1.0" standalone="yes"?>
<Paper uid="H89-2053">
  <Title>Joshi, Aravind K., 1985. How Much Context-Sensitivity is Necessary for Characterizing Structural Descriptions-- Tree Adjoining Grammars. In Dowty, D.; Karttunen, L.; and Zwicky, A. (editors), Natural Language Processing--</Title>
  <Section position="6" start_page="407" end_page="408" type="concl">
    <SectionTitle>
3 CONCLUSION
</SectionTitle>
    <Paragraph position="0"> In 'lexicalized' grammars, each elementary structure is systematically associated with a lexical anchor. These structures specify extended domains of locality (as compared to the domain of locality in CFGs) over which constraints can be stated. The 'grammar' consists of a lexicon in which each lexical item is associated with a finite number of structures for which that item is the anchor.</Paragraph>
    <Paragraph position="1"> Lexicalized grammars suggest a natural two-step parsing strategy. The first step selects the set of structures corresponding to each word in the sentence. The second step tries to combine the selected structures.</Paragraph>
    <Paragraph position="2"> We take Lexicalized TAGs as an instance of lexicalized grammar. We illustrate the organization of the grammar Then we show how the Earley-type parser can take advantage of the two-step parsing strategy. Experimental data show that its performance is thereby drastically improved. The first pass not only filters the grammar used by the parser to produce a relevant subset but also enables the parser to use non-local bottom-up information to guide its search. In Schabes and Joshi (1989) it is also shown that Lexicalization guarantees termination of the parsing algorithm of feature structures for Lexicalized TAGs without a special mechanism such as the use of restrictors.</Paragraph>
    <Paragraph position="3"> The organization of lexicalized grammars, the simplicity and effectiveness of the two-pass strategy (some other technique would perhaps achieve similar results) seem attractive from a linguistic point of view and for processing.</Paragraph>
  </Section>
class="xml-element"></Paper>