<?xml version="1.0" standalone="yes"?>
<Paper uid="A97-1052">
  <Title>Corpus Data TP FP FN</Title>
  <Section position="7" start_page="361" end_page="361" type="concl">
    <SectionTitle>
5 Conclusions and Further Work
</SectionTitle>
    <Paragraph position="0"> The experiment and comparison reported above suggests that our more comprehensive subcategorization class extractor is able both to assign classes to individual verbal predicates and also to rank them according to relative frequency with comparable accuracy to extant systems. We have also demonstrated that a subcategorization dictionary built with the system can improve the accuracy of a probabilistic parser by an appreciable amount.</Paragraph>
    <Paragraph position="1"> The system we have developed is straightforwardly extensible to nominal and adjectival predicates; the existing grammar distinguishes nominal and adjectival arguments from adjuncts structurally, so all that is required is extension of the classifier. Developing an analogous system for another language would be harder but not infeasible; similar taggers and parsers have been developed for a number of languages, but no extant subcategorization dictionaries exist to our knowledge, therefore the lexical statistics we utilize for statistical filtering would need to be estimated, perhaps using the technique described by Brent (1993). However, the entire approach to filtering needs improvement, as evaluation of our results demonstrates that it is the weakest link in our current system.</Paragraph>
    <Paragraph position="2"> Our system needs further refinement to narrow some subcategorization classes, for example, to choose between differing control options with predicative complements. It also needs supplementing with information about diathesis alternation possibilities (e.g. Levin, 1993) and semantic selection preferences on argument heads. Grishman &amp; Sterling (1992), Poznanski &amp; Sanfilippo (1993), Resnik (1993), Ribas (1994) and others have shown that it is possible to acquire selection preferences from (partially) parsed data. Our system already gathers head lemmas in patterns, so any of these approaches could be applied, in principle. In future work, we intend to extend the system in this direction. The ability to recognize that argument slots of different subcategorization classes for the same predicate share semantic restrictions/preferences would assist recognition that the predicate undergoes specific alternations, this in turn assisting inferences about control, equi and raising (e.g. Boguraev &amp; Briscoe, 1987).</Paragraph>
  </Section>
class="xml-element"></Paper>