<?xml version="1.0" standalone="yes"?>
<Paper uid="W00-0715">
  <Title>Experiments on Unsupervised Learning for Extracting Relevant Fragments from Spoken Dialog Corpus</Title>
  <Section position="4" start_page="83" end_page="83" type="metho">
    <SectionTitle>
3 The results of experiments
</SectionTitle>
    <Paragraph position="0"> Below we show examples of labeling and chunking the phrases. As an example we use two phrase from ATIS dialog corpus which includes nearly 20K sentences about flights, reservations, tickets, prices, car rent, flight classes and others.</Paragraph>
  </Section>
  <Section position="5" start_page="83" end_page="84" type="metho">
    <SectionTitle>
WHAT TYPE OF AIRCRAFT IS
USED FOR THIS FLIGHT
</SectionTitle>
    <Paragraph position="0"> quence of tag word labels from the left context and other non tag words on the right up but exclude the next tag word label. Below is the list of the chunks for those two sentences.</Paragraph>
    <Paragraph position="1"> (&lt;what&gt;WHAT) TYPE (&lt;of&gt;OF) AIRCRAFT</Paragraph>
    <Paragraph position="3"> We have divided the corpus into two parts. For each part we did the labeling, chunking and clustering by using Hamming metrics for distance measure. Below we present the words extracted from both parts of the corpus.</Paragraph>
    <Paragraph position="4"> The words extracted from the first 15K sentences. null</Paragraph>
  </Section>
class="xml-element"></Paper>