<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-3240">
  <Title>Learning to Classify Email into &amp;quot;Speech Acts&amp;quot;</Title>
  <Section position="3" start_page="0" end_page="0" type="relat">
    <SectionTitle>
2 Related Work
</SectionTitle>
    <Paragraph position="0"> Our research builds on earlier work defining illocutionary points of speech acts (Searle, 1975), and relating such speech acts to email and workflow tracking (Winograd, 1987, Flores &amp; Ludlow, 1980, Weigant et al, 2003). Winograd suggested that research explicating the speech-act based &amp;quot;language-action perspective&amp;quot; on human communication could be used to build more useful tools for coordinating joint activities. The Coordinator (Winograd, 1987) was one such system, in which users augmented email messages with additional annotations indicating intent.</Paragraph>
    <Paragraph position="1"> While such systems have been useful in limited contexts, they have also been criticized as cumbersome: by forcing users to conform to a particular formal system, they constrain communication and make it less natural (Schoop, 2001); in short, users often prefer unstructured email interactions (Camino et al. 1998). We note that these difficulties are avoided if messages can be automatically annotated by intent, rather than soliciting a statement of intent from the user.</Paragraph>
    <Paragraph position="2"> Murakoshi et al. (1999) proposed an email annotation scheme broadly similar to ours, called a &amp;quot;deliberation tree&amp;quot;, and an algorithm for constructing deliberation trees automatically, but their approach was not quantitatively evaluated.</Paragraph>
    <Paragraph position="3"> The approach is based on recognizing a set of hand-coded linguistic &amp;quot;clues&amp;quot;. A limitation of their approach is that these hand-coded linguistic &amp;quot;clues&amp;quot; are language-specific (and in fact limited to Japanese text.) Prior research on machine learning for text classification has primarily considered classification of documents by topic (Lewis, 1992; Yang, 1999), but also has addressed sentiment detection (Pang et al., 2002; Weibe et al., 2001) and authorship attribution (e.g., Argamon et al, 2003).</Paragraph>
    <Paragraph position="4"> There has been some previous use of machine learning to classify email messages (Cohen 1996; Sahami et al., 1998; Rennie, 2000; Segal &amp; Kephart, 2000). However, to our knowledge, none of these systems has investigated learning methods for assigning email speech acts. Instead, email is generally classified into folders (i.e., according to topic) or according to whether or not it is &amp;quot;spam&amp;quot;. Learning systems have been previously used to automatically detect acts in conversational speech (e.g. Finke et al., 1998).</Paragraph>
  </Section>
class="xml-element"></Paper>