<?xml version="1.0" standalone="yes"?>
<Paper uid="C88-1005">
  <Title>Efficiency Considerations for LFG-Parsers - Incremental and Table-Lookup Techniques</Title>
  <Section position="5" start_page="26" end_page="26" type="concl">
    <SectionTitle>
4 Adequacy and Efficiency of Grammars
</SectionTitle>
    <Paragraph position="0"> LFG-Grammars have been mostly studied from the point of view of linguistic adequacy, i.e. they have been developed in order to cover substantial aspects of natural language syntax phenomena.</Paragraph>
    <Paragraph position="1"> The parser should help the working linguist to find the optimal grammar for a particular language, to test the individual rules of the grammar as well as the general formalism.</Paragraph>
    <Paragraph position="2"> Parsing efficiency can be studied at least at three different levels:  1. in view of the efficiency of developing grammars (concerns the work of the linguist).</Paragraph>
    <Paragraph position="3"> 2. in view o~ optimizing processing of the input grammar (concerus the preproce~sing phase) 3. in view of optimizing the processing of the input sentence&amp;quot;  (concerml the execution phase and the run time system) The present study is concerned mostl~ with the third type of efficiency i.e. ~lith improving parsing efficiency, presupposing linguistic adequm:y of the model. Pructical efficiency of standard parsing technique~ depends on the amount of back tracking and reprocessing needed in cases of errenous analysis. While using standard recursive desct:nt parsing techniques guarantees the comprehensive coverage of the whole language, it does not exploit available information in an optimal manner. Parsing tables with REACH relations entail more information than single phrase structure rules, they acctmtulate information on the distribution of symbols in the whole grannnar and enable the parser to build up the correct c-structure at the first attempt. If for some reason no valid structure is produced, ordinary back tracking applies and the input string is reprocessed.</Paragraph>
    <Paragraph position="4"> In the final version of the paper further details of current improvements will be reported.</Paragraph>
  </Section>
class="xml-element"></Paper>