<?xml version="1.0" standalone="yes"?>
<Paper uid="W99-0310">
  <Title>A recognition-based meta-scheme for dialogue acts annotation</Title>
  <Section position="10" start_page="81" end_page="81" type="concl">
    <SectionTitle>
7 Conclusion
</SectionTitle>
    <Paragraph position="0"> Tag sets are typically developed to respond to specific applications and practical usages, without bothering too much about how the tags themselves relate to the nature of information needed for their assignment in context. This is fine as long as tag sets are assessed in relation to the use they were originally intended for, but much less so if one wants to evaluate the extent to which one tag set translates into another tag set, or to assess the usability of a given tag set for other purposes/applications.</Paragraph>
    <Paragraph position="1"> The multi-dimensional recognition-based meta-scheme described in these pages makes it explicit how intentions relate to the linguistic and contextual information needed for their identification. We showed that this is extremely helpful for scheme comparison, as it sheds light on the precise nature of tag correspondences, well beyond the intuitive grasp provided by tag definitions.</Paragraph>
    <Paragraph position="2"> Preliminary experiments show that a translation of a dialogue tagged with an existing scheme into our meta-scheme is also a useful exercise to assess the internal consistency of the annotated material. If this is confirmed, then use of the meta-scheme should improve scheme design considerably, and should be able to provide procedural and testable guide-lines for dialogue annotators.</Paragraph>
  </Section>
class="xml-element"></Paper>