<?xml version="1.0" standalone="yes"?>
<Paper uid="W03-1019">
  <Title>Investigating Loss Functions and Optimization Methods for Discriminative Learning of Label Sequences</Title>
  <Section position="8" start_page="0" end_page="0" type="concl">
    <SectionTitle>
7 Conclusion and Future Work
</SectionTitle>
    <Paragraph position="0"> In this paper, we investigated how different objective functions and optimization methods affect the accuracy of the sequence labelling task in the discriminative learning framework. Our experiments show that optimizing different objective functions does not have a large affect on the accuracy. Extending the feature space is more effective. We conclude that methods that can use large, possibly infinite number of features may be advantageous over others. We are running experiments where we use a dual formulation of the perceptron algorithm which has the property of being able to use infinitely many features. Our future work includes using SVMs for label sequence learning task.</Paragraph>
  </Section>
class="xml-element"></Paper>