<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-1656">
  <Title>Boosting Unsupervised Relation Extraction by Using NER</Title>
  <Section position="9" start_page="480" end_page="480" type="concl">
    <SectionTitle>
5 Conclusions
</SectionTitle>
    <Paragraph position="0"> We have presented the URES system for autonomously extracting relations from the Web. We showed how to improve the precision of the system by classifying the extracted instances using the properties of the patterns and sentences that generated the instances and how to utilize a simple NER component. The cross-predicate tests showed that classifier that performs well for all relations can be built using a small amount of labeled data for any particular relation. We performed an experimental comparison between URES, URES-NER and the state-of-the-art KnowItAll system, and showed that URES can double or even triple the recall achieved by KnowItAll for relatively rare relation instances, and get an additional 5-15% boost in recall by utilizing a simple NER. In particular we have shown that URES is more effective in identifying low-frequency instances, due to its more expressive rule representation, and its classifier (augmented by NER) that inhibits those rules from overgeneralizing.</Paragraph>
  </Section>
class="xml-element"></Paper>