<?xml version="1.0" standalone="yes"?>
<Paper uid="N06-2025">
  <Title>Syntactic Kernels for Natural Language Learning: the Semantic Role Labeling Case</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Recently, several tree kernels have been applied to natural language learning, e.g. (Collins and Duffy, 2002; Zelenko et al., 2003; Cumby and Roth, 2003; Culotta and Sorensen, 2004; Moschitti, 2004). Despite their promising results, three general objections against kernel methods are raised: (1) only a subset of the dual space features are relevant, thus, it may be possible to design features in the primal space that produce the same accuracy with a faster computation time; (2) in some cases the high number of features (substructures) of the dual space can produce overfitting with a consequent accuracy decrease (Cumby and Roth, 2003); and (3) the computation time of kernel functions may be too high and prevent their application in real scenarios.</Paragraph>
    <Paragraph position="1"> In this paper, we study the impact of the sub-tree (ST) (Vishwanathan and Smola, 2002), subset tree (SST) (Collins and Duffy, 2002) and partial tree (PT) kernels on Semantic Role Labeling (SRL). The PT kernel is a new function that we have designed to generate larger substructure spaces. Moreover, to solve the computation problems, we propose algorithms which evaluate the above kernels in linear average running time.</Paragraph>
    <Paragraph position="2"> We experimented such kernels with Support Vector Machines (SVMs) on the classification of semantic roles of PropBank (Kingsbury and Palmer, 2002) and FrameNet (Fillmore, 1982) data sets. The results show that: (1) the kernel approach provides the same accuracy of the manually designed features.</Paragraph>
    <Paragraph position="3"> (2) The overfitting problem does not occur although the richer space of PTs does not provide better accuracy than the one based on SST. (3) The average running time of our tree kernel computation is linear.</Paragraph>
    <Paragraph position="4"> In the remainder of this paper, Section 2 introduces the different tree kernel spaces. Section 3 describes the kernel functions and our fast algorithms for their evaluation. Section 4 shows the comparative performance in terms of execution time and accuracy. null</Paragraph>
  </Section>
class="xml-element"></Paper>