<?xml version="1.0" standalone="yes"?>
<Paper uid="W02-2001">
  <Title>Extracting the Unextractable: A Case Study on Verb-particles</Title>
  <Section position="8" start_page="0" end_page="0" type="concl">
    <SectionTitle>
8 Conclusion
</SectionTitle>
    <Paragraph position="0"> In conclusion, this paper has been concerned with the extraction of English verb{particle constructions from raw text corpora. Three basic methods were proposed, based on tagger output, chunker output and a chunk grammar; the chunk grammar method was optionally combined with attachment resolution to determine the syntactic structure of verb{preposition pairs in ambiguous constructs. We then experimented with combining the output of the three methods together into a single classifler, and further complemented the feature space with a number of lexical and frequentistic features, culminating in an F-score of 0.865 over the WSJ.</Paragraph>
    <Paragraph position="1"> It is relatively simple to adapt the methods described here to output subcategorisation types, rather than a binary judgement on verb{ particlehood. This would allow the extracted output to be fed directly into the LinGO-ERG for use in parsing. We are also interested in extending the method to extract prepositional verbs, many of which appear in the attachment resolution data and are subsequently flltered out by the consolidated classifler.</Paragraph>
  </Section>
class="xml-element"></Paper>