<?xml version="1.0" standalone="yes"?>
<Paper uid="W99-0310">
  <Title>A recognition-based meta-scheme for dialogue acts annotation</Title>
  <Section position="9" start_page="81" end_page="81" type="evalu">
    <SectionTitle>
6 Annotation and meta-scheme
</SectionTitle>
    <Paragraph position="0"> As already pointed out above, the meta-scheme proposed here does not per se fulfill some important prerequisites for an annotation scheme. It is useful, at this stage, to elaborate this point. First, multidimensionality and orthogonality of the assumed multiple dimensions seem to be operationally cumbersome and, in general, detract from reliability in actual tagging practice. Furthermore, in the meta-scheme all classificatory dimensions are conceived of as being on a par. This means that we deliberately make no assumption as to what dimension of annotation ultimately provides information about the primary intended illocutionary act of an utterance, and how information along one dimension relates to information encoded at another dimension.</Paragraph>
    <Paragraph position="1"> This is not very informative from the point of view of annotation, but represents a very useful feature for scheme customization, as it makes it possible to modify/adapt an existing annotation scheme by collapsing some analytical dimensions in a controlled way.</Paragraph>
    <Paragraph position="2"> Finally, it should be appreciated that the list of dimensions provided here is not meant to be either exhaustive or minimal, in the sense that every tag should be classified along each dimension. Other possible dimensions of analysis can include, for example, kinesic information, to account for dialogue acts performed through non-verbal communicative behavior, such as nodding, smiling and pointing.</Paragraph>
    <Paragraph position="3"> As long as dimensions are rigorously defined, this should clarify the intended use of a scheme considerably. null</Paragraph>
  </Section>
class="xml-element"></Paper>