<?xml version="1.0" standalone="yes"?>
<Paper uid="W05-1002">
  <Title>Verb subcategorization kernels for automatic semantic labeling</Title>
  <Section position="3" start_page="0" end_page="10" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Some theories of verb meaning are based on syntactic properties, e.g. the alternations of verb arguments (Levin, 1993). In turn, Verb Subcategorization Frame (SCF) characterizes different syntactic alternations, thus, it plays a central role in the linking theory between verb semantics and their syntactic structures.</Paragraph>
    <Paragraph position="1"> Figure 1 shows the parse tree for the sentence &amp;quot;John rented a room in Boston&amp;quot; along  with the semantic shallow information embodied by the verbal predicate to rent and its three arguments: Arg0, Arg1 and ArgM. The SCF of such verb, i.e.</Paragraph>
    <Paragraph position="2"> NP-PP, provides a synthesis of the predicate argument structure.</Paragraph>
    <Paragraph position="3"> Currently, the systems which aim to derive semantic shallow information from texts recognize the SCF of a target verb and represent it as a flat feature (e.g. (Xue and Palmer, 2004; Pradhan et al., 2004)) in the learning algorithm. To achieve this goal, a lexicon which describes the SCFs for each verb, is required. Such a resource is difficult to find especially for specific domains, thus, several methods to automatically extract SCF have been proposed (Korhonen, 2003). In (Moschitti, 2004), an alternative to the SCF extraction was proposed, i.e. the SCF kernel (SK). The subcategorization frame of verbs was implicitly represented by means of the syntactic sub-trees which include the predicate with its arguments. The similarity between such syntactic structures was evaluated by means of convolution kernels.</Paragraph>
    <Paragraph position="4"> Convolution kernels are machine learning approaches which aim to describe structured data in  terms of its substructures. The similarity between two structures is carried out by kernel functions which determine the number of common substructures without evaluating the overall substructure space. Thus, if we associate two SCFs with two subtrees, we can measure their similarity with such functions applied to the two trees. This approach determines a more syntactically motivated verb partition than the traditional method based on flat SCF representations (e.g. the NP-PP of Figure 1). The subtrees associated with SCF group the verbs which have similar syntactic realizations, in turn, according to Levin's theories, this would suggest that they are semantically related.</Paragraph>
    <Paragraph position="5"> A preliminary study on the benefit of such kernels was measured on the classification accuracy of semantic arguments in (Moschitti, 2004). In such work, the improvement on the PropBank arguments (Kingsbury and Palmer, 2002) classification suggests that SK adds information to the prediction of semantic structures. On the contrary, the performance decrease on the FrameNet data classification shows the limit of such approach, i.e. when the syntactic structures are shared among several semantic roles SK seems to be useless.</Paragraph>
    <Paragraph position="6"> In this article, we use Support Vector Machines (SVMs) to deeply analyze the role of SK in the automatic predicate argument classification. The major novelty of the article relates to the extensive experimentation carried out on the PropBank (Kingsbury and Palmer, 2002) and FrameNet (Fillmore, 1982) corpora with diverse levels of task complexity, e.g. test instances of unseen predicates (typical of free-text processing). The results show that: (1) once a structural representation of a linguistic object, e.g. SCF, is available we can use convolution kernels to study its connections with another linguistic phenomenon, e.g. the semantic predicate arguments. (2) The tree kernels automatically derive the features (structures) which support also a sort of back-off estimation in case of unseen verbs. (3) The structural features are in general robust in all testing conditions.</Paragraph>
    <Paragraph position="7"> The remainder of this article is organized as follows: Section 2 defines the Predicate Argument Extraction problem and the standard solution to solve it. In Section 3 we present our kernels whereas in Section 4 we show comparative results among SVMs using standard features and the proposed kernels. Finally, Section 5 summarizes the conclusions.</Paragraph>
  </Section>
class="xml-element"></Paper>