<?xml version="1.0" standalone="yes"?>
<Paper uid="C04-1008">
  <Title>Annotating and measuring temporal relations in texts</Title>
  <Section position="6" start_page="0" end_page="0" type="concl">
    <SectionTitle>
7 Conclusion
</SectionTitle>
    <Paragraph position="0"> The aim of this study was to show the feasibility of annotating temporal relations in a text and to propose a methodology for the task. We thus define a way of evaluating the results, abstracting away from variations of human descriptions for similar temporal situations. Our preliminary results seem promising in this respect. Obviously, parts of the method need some polishing, and we need to extend the study to a larger data set. It remains to be seen how improving part of speech tagging, syntactic analysis and discourse modeling can influence the outcome of the task. Specifically, some work needs to be done to evaluate the detection of temporal adjuncts, a major source of information in the process.</Paragraph>
    <Paragraph position="1"> We could also try to mix our symbolic method with some empirical learning. Provided we can collect more annotated data, it would be easy to improve the discourse model by (at least local) optimization on the space of possible rules, starting with our own set. We hope that the measures of temporal information we have used will help in all these aspects, but we are also planning to further investigate their properties and that of other candidate measures not considered here.</Paragraph>
  </Section>
class="xml-element"></Paper>