<?xml version="1.0" standalone="yes"?>
<Paper uid="W97-1003">
  <Title>A Preliminary Study of Word Clustering Based on Syntactic Behavior</Title>
  <Section position="7" start_page="0" end_page="0" type="concl">
    <SectionTitle>
8 Conclusion
</SectionTitle>
    <Paragraph position="0"> We have presented a method which constructs classes of words with similar syntactic behavior, or binary trees that reflect word similarity, by clustering words using treebank data. In this way it is possible to discover particular types of behavior, such as the peculiar behavior of the gerund including, verbs that modify an entire clause (raising verbs), nouns that prefer either subject position or object position, or prepositions that prefer locative phrases.</Paragraph>
    <Paragraph position="1"> Most of the classes found in this way would not be found if clustering were performed on the basis of co-occurrences, as has been described in the literature. For example, the verbs \[tend, plan, continue, want, need, seem, appear\] share a particular sentence structure rather than, say, the sort of noun that becomes the object.</Paragraph>
    <Paragraph position="2"> As became clear from the case study of prepositions, the clustering process reveals similarities in the syntactic structure in which words appear which in some cases can be clearly felt by intuition. For example, the words in, on and at often are the head of locative prepositional phrases, and a preposition such as within usually is the head of a temporal prepositional phrase. Using this method these intuitions can be quantified.</Paragraph>
    <Paragraph position="3"> One of the applications we described is that of a decision-tree based system for syntactic analysis.</Paragraph>
    <Paragraph position="4"> We are currently applying the method in this application, which will be described in later publications.</Paragraph>
  </Section>
class="xml-element"></Paper>