<?xml version="1.0" standalone="yes"?>
<Paper uid="W03-0418">
  <Title>Identifying Events using Similarity and Context</Title>
  <Section position="6" start_page="0" end_page="0" type="evalu">
    <SectionTitle>
5 Related Work
</SectionTitle>
    <Paragraph position="0"> Two lines of research are relevant to this work. First, our research is based on research into scripts. Second, recent work in semantic lexicon learning is similar to our work, although it focuses on learning related words, not related clauses. In addition, extraction patterns and case frames bear some resemblance to our events. Riloff and Schmelzenbach's work (1998) is an example of this line of research. However, events use contextual information about other events, unlike extraction patterns and case frames.</Paragraph>
    <Paragraph position="1"> The idea of a script originated with Schank and Abelson (1977) through research on human knowledge structures, and was demonstrated in the SAM system (Cullingford, 1978). Later work includes manual creation of a variety of knowledge structures including scripts to understand stories (Lehnert et al., 1983), application of manually generated scripts to the processing of newswire stories (DeJong, 1982), and a combination of applying manually generated scripts to information retrieval and applying genetic algorithms to adjusting existing scripts (Mauldin, 1989).</Paragraph>
    <Paragraph position="2"> Semantic lexicons have been the focus of much research. WordNet (Fellbaum, 1998) is a prominent example of a manually generated lexicon. Two recent projects in learning semantic lexicons apply automated techniques to a small set of human provided seed words to create lists of words that the systems assign to the same semantic category. Each project uses a different technique to evaluate word similarity. Thelen (2002) uses similar context within a sentence. Phillips (2002) mines syntactic structures. Other researchers have also clustered words to create semantic lexicons. Lin (1998) created a thesaurus using syntactic relationships with other words.</Paragraph>
    <Paragraph position="3"> Rooth et al. (1999) used clustering to create clusters similar to Levin verb classes (Levin, 1993). Pereira, Tishby and Lee (1993) clustered words according to context.</Paragraph>
  </Section>
class="xml-element"></Paper>