<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-1113">
  <Title>Using Synonym Relations In Chinese Collocation Extraction</Title>
  <Section position="5" start_page="178" end_page="178" type="concl">
    <SectionTitle>
5 Conclusion and Further Work
</SectionTitle>
    <Paragraph position="0"> In this paper, we have presented a method to extract bi-gram collocations using lexical statistics model with synonyms information. Our method reaches the precision rate of 43% for the tested data. Comparing to the precision of 30% using lexical statistics only, our improvement is close to 50%. In additional, the recall improved 30%. The contribution is that we have made use of synonym information which is plentiful in the natural language use and it works well to supplement the shortcomings of lexical statistical method.</Paragraph>
    <Paragraph position="1"> Manning claimed that the lack of valid substitution for a synonym is a characteristics of collocations in general (Manning and Schutze 1999). To extend our work, we consider the use of synonym information can be further applied to help identify collocations of different types.</Paragraph>
    <Paragraph position="2"> Our preliminary study has suggested that collocation can be classified into 4 types: Type 0 Collocation: Fully fixed collocation which include some idioms, proverbs and sayings such as &amp;quot; &amp;quot; &amp;quot; &amp;quot; and so on.</Paragraph>
    <Paragraph position="3"> Type 1 Collocation: Fixed collocation in which the appearance of one word implies the co-occurrence of another one such as &amp;quot; &amp;quot;. Type 2 Collocation: Strong collocation which allows very limited substitution of the components, for example, &amp;quot; &amp;quot;, &amp;quot; &amp;quot;, &amp;quot; &amp;quot; and so on.</Paragraph>
    <Paragraph position="4"> Type 3 Collocation: Normal collocation which allows more substitution of the components, however a limitation is still required. For example,</Paragraph>
    <Paragraph position="6"> By using synonym information and define substitutability, we can validate whether collocations are fixed collocations, strong collocations with very limited substitutions, or general collocations that can be substituted more freely.</Paragraph>
  </Section>
class="xml-element"></Paper>