<?xml version="1.0" standalone="yes"?>
<Paper uid="C04-1147">
  <Title>Fast Computation of Lexical Affinity Models</Title>
  <Section position="6" start_page="0" end_page="0" type="concl">
    <SectionTitle>
5 Conclusion
</SectionTitle>
    <Paragraph position="0"> We presented a framework for the fast and effective computation of lexical affinity models. Instead of using arbitrary windows to compute word similarity measures, we model lexical affinity using the complete observed distance distribution along with independence and parametric models for this distribution. Our results shows that, with minimal effort to adapt the models, we achieve good results by applying this framework to simple natural language tasks, such as TOEFL synonym questions and GRE fill-in-the-blanks tests. This framework allows the use of terabyte-scale corpora by providing a fast algorithm to extract pairs of co-occurrence for the models, thus enabling the use of more precise estimators. null</Paragraph>
  </Section>
class="xml-element"></Paper>