<?xml version="1.0" standalone="yes"?>
<Paper uid="P06-1128">
  <Title>Semantic Retrieval for the Accurate Identification of Relational Concepts in Massive Textbases Yusuke Miyao[?] Tomoko Ohta[?] Katsuya Masuda[?] Yoshimasa Tsuruoka+</Title>
  <Section position="7" start_page="1023" end_page="1023" type="concl">
    <SectionTitle>
5 Conclusion
</SectionTitle>
    <Paragraph position="0"> We demonstrated a text retrieval system for MEDLINE that exploits pre-computed semantic annotations5. Experimental results revealed that the proposed system is sufficiently efficient for real-time text retrieval and that the precision of retrieval was remarkably high. Analysis of residual errors showed that the handling of noun phrase structures and the improvement of term recognition will increase retrieval accuracy. Although the present paper focused on MEDLINE, the NLP tools used in this system are domain/task independent. This framework will thus be applicable to other domains such as patent documents.</Paragraph>
    <Paragraph position="1"> The present framework does not conflict with conventional IR/IE techniques, and integration with these techniques is expected to improve the accuracy and usability of the proposed system. For example, query expansion and relevancy feedback can be integrated in a straightforward way in order to improve accuracy. Document ranking is useful for the readability of retrieved results. IE systems can be applied off-line, in the manner of the deep parser in our system, for annotating sentences with target information of IE. Such annotations will enable us to retrieve higher-level concepts, such as relationships among relational concepts.</Paragraph>
  </Section>
class="xml-element"></Paper>