<?xml version="1.0" standalone="yes"?>
<Paper uid="P04-1080">
  <Title>Learning Word Senses With Feature Selection and Order Identification Capabilities</Title>
  <Section position="6" start_page="0" end_page="0" type="relat">
    <SectionTitle>
4 Related Work
</SectionTitle>
    <Paragraph position="0"> Besides the two works (Pantel and Lin, 2002; Sch&amp;quot;utze, 1998), there are other related efforts on word sense discrimination (Dorow and Widdows, 2003; Fukumoto and Suzuki, 1999; Pedersen and Bruce, 1997).</Paragraph>
    <Paragraph position="1"> In (Pedersen and Bruce, 1997), they described an experimental comparison of three clustering algorithms for word sense discrimination. Their feature sets included morphology of target word, part of speech of contextual words, absence or presence of particular contextual words, and collocation of fre- null horizontal axis corresponds to the context window size. Solid line represents the result of FSGMM + binary, dashed line denotes the result of CGDSVD + idf, and dotted line is the result of CGDterm + idf. Square marker denotes '2 based feature ranking, while cross marker denotes freq based feature ranking.</Paragraph>
    <Paragraph position="2">  datasets.</Paragraph>
    <Paragraph position="3"> quent words. Then occurrences of target word were grouped into a pre-defined number of clusters. Similar with many other algorithms, their algorithm also required the cluster number to be provided.</Paragraph>
    <Paragraph position="4"> In (Fukumoto and Suzuki, 1999), a term weight learning algorithm was proposed for verb sense disambiguation, which can automatically extract nouns co-occurring with verbs and identify the number of senses of an ambiguous verb. The weakness of their method is to assume that nouns co-occurring with verbs are disambiguated in advance and the number of senses of target verb is no less than two.</Paragraph>
    <Paragraph position="5"> The algorithm in (Dorow and Widdows, 2003) represented target noun word, its neighbors and their relationships using a graph in which each node denoted a noun and two nodes had an edge between them if they co-occurred with more than a given number of times. Then senses of target word were iteratively learned by clustering the local graph of similar words around target word. Their algorithm required a threshold as input, which controlled the number of senses.</Paragraph>
  </Section>
class="xml-element"></Paper>