<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-1315">
  <Title>Sydney, July 2006. c(c)2006 Association for Computational Linguistics Empirical Verification of Adjacency Pairs Using Dialogue Segmentation</Title>
  <Section position="10" start_page="105" end_page="106" type="concl">
    <SectionTitle>
6 Conclusions and Future Work
</SectionTitle>
    <Paragraph position="0"> We can draw some tentative conclusions from this work. First of all, the dialogue segmentation combined with the kh  test for significance yields information about what is likely to happen, not just for the next utterance, but somewhere in the next chunk. This will help to overcome the limitations imposed by the traditional 'previous  tag' feature. We are working to implement this information into a model where the expectations inherent in a first pair part are retained when not immediately fulfilled. The expectations will also decay with time.</Paragraph>
    <Paragraph position="1"> Second, this approach provides empirical evidence for adjacency pairs mentioned in the literature on conversation analysis. The noise reduction feature of the kh  test gives more weight to legitimate adjacency pairs where they appear in the data.</Paragraph>
    <Paragraph position="2"> An intriguing possibility for the chunked data is that of chunk matching. Nearest-neighbour algorithms are already used for classification tasks (including DA tagging for individual utterances), but once segmented, the dialogue chunks could be compared against each other as a classification tool as in a nearest-neighbour algorithm.</Paragraph>
  </Section>
class="xml-element"></Paper>