<?xml version="1.0" standalone="yes"?>
<Paper uid="J01-4003">
  <Title>A Corpus-Based Evaluation of Centering and Pronoun Resolution</Title>
  <Section position="5" start_page="517" end_page="518" type="metho">
    <SectionTitle>
6. Conclusions
</SectionTitle>
    <Paragraph position="0"> In this paper we first presented a new pronoun resolution algorithm, Left-Right Centering, which adheres to the constraints of centering theory and was inspired by the need to remedy a lack of incremental processing in Brennan, Friedman, and Pollard's (1987) method. Second, we compared LRC's performance with that of three other leading pronoun resolution algorithms, each one restricted to using only syntactic information. This comparison is significant in its own right because these algorithms have not been previously compared, in computer-encoded form, on a common corpus. Coding  Tetreault Centering and Pronoun Resolution all the algorithms allows one to quickly test them on a large corpus and eliminates human error. Third, we tried to improve LRC's performance by incorporating theories on Cf-list construction derived from psycholinguistic experiments. Our corpus-based evaluation showed that prepended phrases should not be ranked prominently in the Cf-list as Gordon, Grosz, and Gilliom (1993) suggest. Our results also showed that Walker and Prince's (1996) complex NP assumption performs marginally better than the opposite theory based on experimental results. We believe that corpus-based analyses such as this one not only increase performance in resolution algorithms but also can aid in validating the results of psycholinguistic studies, which are usually based on small sequences of utterances.</Paragraph>
  </Section>
class="xml-element"></Paper>