<?xml version="1.0" standalone="yes"?>
<Paper uid="P05-1006">
  <Title>The Role of Semantic Roles in Disambiguating Verb Senses</Title>
  <Section position="6" start_page="47" end_page="48" type="relat">
    <SectionTitle>
4 Related Work
</SectionTitle>
    <Paragraph position="0"> Our approach of using rich linguistic features combined in a single maximum entropy framework contrasts with that of (Florian et al., 2002). Their feature space was much like ours, but did not include semantic class features for noun complements. With this more impoverished feature set, they experimented with combining diverse classifiers to achieve an improvement of 2.1% over all parts of speech (noun, verb, adjective) in the Senseval-2 lexical sample task; however, this improvement was over an initial accuracy of 56.6% on verbs, indicating that their performance is still below ours for verbs.</Paragraph>
    <Paragraph position="1"> (Lee and Ng, 2002) explored the relative contribution of different knowledge sources and learning algorithms to WSD; they used Support Vector Machines (SVM) and included local collocations and syntactic relations, and also found that adding syntactic features improved accuracy. Our features are similar to theirs, but we added semantic class features for the verb arguments. We found that the difference in machine learning algorithms did not play a large role in performance; when we used our features in SVM we obtained almost no difference in performance over using maximum entropy models with Gaussian priors.</Paragraph>
    <Paragraph position="2"> (Gomez, 2001) described an algorithm using WordNet to simultaneously determine verb senses and attachments of prepositional phrases, and iden- null tify thematic roles and adjuncts; our work is different in that it is trained on manually annotated corpora to show the relevance of semantic roles for verb sense disambiguation.</Paragraph>
  </Section>
class="xml-element"></Paper>