<?xml version="1.0" standalone="yes"?>
<Paper uid="P05-1072">
  <Title>Semantic Role Labeling Using Different Syntactic Views[?]</Title>
  <Section position="3" start_page="581" end_page="581" type="intro">
    <SectionTitle>
2 Baseline System
</SectionTitle>
    <Paragraph position="0"> For our experiments, we use Feb 2004 release of PropBank1 (Kingsbury and Palmer, 2002; Palmer et al., 2005), a corpus in which predicate argument relations are marked for verbs in the Wall Street Journal (WSJ) part of the Penn TreeBank (Marcus et al., 1994). PropBank was constructed by assigning semantic arguments to constituents of hand-corrected TreeBank parses. Arguments of a verb are labeled ARG0 to ARG5, where ARG0 is the PROTO-AGENT, ARG1 is the PROTO-PATIENT, etc.</Paragraph>
    <Paragraph position="1"> In addition to these CORE ARGUMENTS, additional ADJUNCTIVE ARGUMENTS, referred to as ARGMs are also marked. Some examples are ARGM-LOC, for locatives; ARGM-TMP, for temporals; ARGM-MNR, for manner, etc. Figure 1 shows a syntax tree along with the argument labels for an example extracted from PropBank. We use Sections 02-21 for training, Section 00 for development and Section 23 for testing.</Paragraph>
    <Paragraph position="2"> We formulate the semantic labeling problem as a multi-class classification problem using Support Vector Machine (SVM) classifier (Hacioglu et al., 2003; Pradhan et al., 2003; Pradhan et al., 2004) TinySVM2 along with YamCha3 (Kudo and Mat- null to implement the system. Using what is known as the ONE VS ALL classification strategy, n binary classifiers are trained, where n is number of semantic classes including a NULL class.</Paragraph>
    <Paragraph position="3"> The baseline feature set is a combination of features introduced by Gildea and Jurafsky (2002) and ones proposed in Pradhan et al., (2004), Surdeanu et al., (2003) and the syntactic-frame feature proposed in (Xue and Palmer, 2004). Table 1 lists the features used.</Paragraph>
  </Section>
class="xml-element"></Paper>