<?xml version="1.0" standalone="yes"?>
<Paper uid="P98-1011">
  <Title>Evaluating a Focus-Based Approach to Anaphora Resolution*</Title>
  <Section position="8" start_page="77" end_page="77" type="concl">
    <SectionTitle>
7 Conclusion
</SectionTitle>
    <Paragraph position="0"> A focus-based approach to pronoun resolution has been implemented within the LaSIE IE system and evaluated on real-world texts. The results show no significant preformance increase over a simpler heuristic-based approach. The main limitation of the focus-based approach is its reliance on a robust syntactic/semantic analysis to find the focus on which all the IRs depend. Examining performance on the real-world data also raises questions about the theoretical assumptions of focus-based approaches, in particular whether focus is always a favoured antecedent, or whether this depends, to some extent, on discourse style.</Paragraph>
    <Paragraph position="1"> Analysing the differences in the results of the focus- and non-focus-based approaches, does show that the focus-based rules are commonly required when the simple syntactic and semantic rules propose a set of equivalent antecedents and can only select, say, the closest arbitrarily. A combined approach is therefore suggested, but whether this would be more effective than further refining the resolution rules of the focus-based approach, or improving parse results and adding more detailed semantic constraints, remains an open question.</Paragraph>
  </Section>
class="xml-element"></Paper>