<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-3402">
  <Title>Off-Topic Detection in Conversational Telephone Speech</Title>
  <Section position="1" start_page="0" end_page="0" type="abstr">
    <SectionTitle>
Abstract
</SectionTitle>
    <Paragraph position="0"> In a context where information retrieval is extended to spoken &amp;quot;documents&amp;quot; including conversations, it will be important to provide users with the ability to seek informational content, rather than socially motivated small talk that appears in many conversational sources. In this paper we present a preliminary study aimed at automatically identifying &amp;quot;irrelevance&amp;quot; in the domain of telephone conversations. We apply a standard machine learning algorithm to build a classifier that detects off-topic sections with better-than-chance accuracy and that begins to provide insight into the relative importance of features for identifying utterances as on topic or not.</Paragraph>
  </Section>
class="xml-element"></Paper>