<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-0705">
  <Title>Applying Coreference to Improve Name Recognition</Title>
  <Section position="11" start_page="99" end_page="99" type="concl">
    <SectionTitle>
9 Conclusions and Future Work
</SectionTitle>
    <Paragraph position="0"> In this paper, we presented a novel idea of applying coreference information to improve name recognition. We used both a statistical filter based on a set of coreference features and rules for correcting specific errors in name recognition.</Paragraph>
    <Paragraph position="1"> Overall, we obtained an absolute improvement of 3.1% in F score. Put another way, we were able to eliminate about 60% of erroneous name tags with only a small loss in recall.</Paragraph>
    <Paragraph position="2"> The methods were tested on a Chinese name tagger, but most of the techniques should be applicable to other languages. More generally, it offers an example of using global and cross-document information to improve local decisions for information extraction. Such methods will be important for breaking the 'performance ceiling' in many areas of information extraction.</Paragraph>
    <Paragraph position="3"> In the future, we plan to experiment with improvements in coreference resolution (in particular, adding pronoun resolution) to see if we can obtain further gains in name recognition. We also intend to explore the production of multiple tagging hypotheses by our statistical name tagger, with the alternative hypotheses then reranked using global information. This may allow us to replace some of our hand-coded error-correction rules with corpus-trained methods.</Paragraph>
  </Section>
class="xml-element"></Paper>