<?xml version="1.0" standalone="yes"?>
<Paper uid="W97-0813">
  <Title>Inferring Semantic Similarity from Distributional Evidence: an Analogy-based Approach to Word Sense Disambiguation*</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Sense disambiguation of a given word occurrence in a specific context (hereafter WSD) requires appeal to a wide typology of cues, ranging from syntactic subcategorization to lexico-semantic information and subject domain. In this paper we will focus on the use of lexico-semantic information, and will try to tackle the related problem of measuring the semantic similarity between the surrounding context of the word to be disarnbiguated and typical patterns of use of that word in a dictionary database. In the literature, semantic similarity is usually assessed with reference to a hierarchically structured thesaurus (e.g. WordNet, \[Miller, 1990\]). The goal of the paper is to investigate an alternative way of measuring semantic similarity, based on distributional evidence, and to show that this evidence can reliably be used to disambiguate words in context. To this end, we will make use of textual and lexical resources of Italian: nonetheless we are convinced that the general point made in this paper has a cross-linguistic validity.</Paragraph>
  </Section>
class="xml-element"></Paper>