<?xml version="1.0" standalone="yes"?>
<Paper uid="N06-1042">
  <Title>Learning Morphological Disambiguation Rules for Turkish</Title>
  <Section position="7" start_page="332" end_page="332" type="concl">
    <SectionTitle>
5 Contributions
</SectionTitle>
    <Paragraph position="0"> We have presented an automated approach to learn morphological disambiguation rules for Turkish using a novel decision list induction algorithm, GPA.</Paragraph>
    <Paragraph position="1"> The only input to the rules are the surface attributes of a ve word window. The approach can be generalized to other agglutinative languages which share the common challenge of a large number of potential tags. Our approach for resolving the data sparseness problem caused by the large number of tags is to generate a separate model for each morphological feature. The predictions for individual features are probabilistically combined based on the accuracy of each model to select the best tag. We were able to achieve an accuracy around 96% using this approach.</Paragraph>
  </Section>
class="xml-element"></Paper>