<?xml version="1.0" standalone="yes"?>
<Paper uid="P98-1051">
  <Title>Error Driven Word Sense Disambiguation</Title>
  <Section position="4" start_page="0" end_page="320" type="intro">
    <SectionTitle>
2 The Resources
</SectionTitle>
    <Paragraph position="0"> We decided to take advantage of the syntactic structures already contained in the Penn Tree Bank (PTB) (Mitchell et al., 1995) in order to build a large set of functional relation pairs (much as in Resnik (1997)). These relations are then used to learn how to perform semantic disambiguation. To distinguish word meanings we use the top 45 semantic tags included in Word-Net (Miller, 1990). The non-supervised Brill algorithm is used to learn and then to apply semantic disambiguation rules. The semantically hand-tagged Brown Corpus is used to evaluate the performance of automatically acquired rules.</Paragraph>
    <Section position="1" start_page="0" end_page="320" type="sub_section">
      <SectionTitle>
2.1 Obtaining Functional Structures.
</SectionTitle>
      <Paragraph position="0"> We consider as crucial for semantic disambiguation the following functional relations: SUB J/VERB, VERB/OBJ, VERB/PREP/PREP null OBJ, NOUN/PREP/PREP-OBJ.</Paragraph>
      <Paragraph position="1"> In order to extract them, we parsed the PTB structures using Zebu (Laubusch, 1994), a LARLR(1) parser implemented in LISP. The parser scans the trees, collecting information about relevant functional relations and writing them out in an explicit format. For instance, the fragment you do something to the economy, after some intermediate steps which are described in Dini et al. (1998a) and Dini et al. (1998b), is  transformed into: HASOBJ do something HASSBJ do you PREPMOD do TO economy</Paragraph>
    </Section>
    <Section position="2" start_page="320" end_page="320" type="sub_section">
      <SectionTitle>
2.2 Adding Lexical Semantics.
</SectionTitle>
      <Paragraph position="0"> The WordNet team has developed a general semantic tagging scheme where every set of synonymous senses, synsets, is tagged with one of 45 tags as in WordNet version 1.5. We use these tags to label all the content words contained in extracted functional relations. We associate each word with all its possible senses ordered in a canonical way. The semantically tagged version of the sample sentence given above is:</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>