<?xml version="1.0" standalone="yes"?>
<Paper uid="P97-1011">
  <Title>Learning Features that Predict Cue Usage</Title>
  <Section position="6" start_page="85" end_page="85" type="concl">
    <SectionTitle>
5 Discussion and Conclusions
</SectionTitle>
    <Paragraph position="0"> We have presented the results of machine learning experiments concerning cue occurrence and placement.</Paragraph>
    <Paragraph position="1"> As (Litman, 1996) observes, this sort of empirical work supports the utility of machine learning techniques applied to coded corpora. As our study shows, individual features have no predictive power for cue occurrence. Moreover, it is hard to see how the best combination of individual features could be found by manual inspection.</Paragraph>
    <Paragraph position="2"> Our results also provide guidance for those building text generation systems. This study clearly indicates that segment structure, most notably the ordering of core and contributor, is crucial for determining cuc occurrence. Recall that it was only by considering Corel and Core~ relations in distinct datasets that we were able to obtain perspicuous decision trees that signifcantly reduce the error rate.</Paragraph>
    <Paragraph position="3"> This indicates that the representations produced by discourse planners should distinguish those elements that constitute the core of each discourse segment, in addition to representing the hierarchical structure of segments. Note that the notion of core is related to the notions of nucleus in RST, intended effect in (Young and Moore, 1994), and of point of a move in (Elhadad and McKeown, 1990), and that text generators representing these notions exist.</Paragraph>
    <Paragraph position="4"> Moreover, in order to use the decision trees derived here, decisions about whether or not to make the core explicit and how to order the core and contributor(s) must be made before deciding cue occurrence, e.g., by exploiting other factors such as focus (McKeown, 1985) and a discourse history.</Paragraph>
    <Paragraph position="5"> Once decisions about core:contributor ordering and cuc occurrence have been made, a generator must still determine where to place cues and select appropriate Icxical items. A major focus of our future research is to explore the relationship between the selection and placement decisions. Elsewhere, we have found that particular lexical items tend to have a preferred location, defined in terms of functional (i.e., core or contributor) and linear (i.e., first or second relatum) criteria (Moser and Moore, 1997). Thus, if a generator uses decision trees such as the one shown in Figure 3 to determine where a cuc should bc placed, it can then select an appropriate cue from those that can mark the given intentional / informational relations, and are usually placed in that functional-linear location. To evaluate this strategy, we must do further work to understand whether there are important distinctions among cues (e.g., so, because) apart from their different preferred locations. The work of Elhadad (1990) and Knott (1996) will help in answering this question.</Paragraph>
    <Paragraph position="6"> Future work comprises further probing into machine learning techniques, in particular investigating whether other learning algorithms are more appropriate for our problem (Mooney, 1996), especially algorithms that take into account some a priori knowledge about features and their dependencies.</Paragraph>
  </Section>
class="xml-element"></Paper>