<?xml version="1.0" standalone="yes"?>
<Paper uid="H05-1118">
  <Title>Proceedings of Human Language Technology Conference and Conference on Empirical Methods in Natural Language Processing (HLT/EMNLP), pages 939-946, Vancouver, October 2005. c(c)2005 Association for Computational Linguistics Integrating linguistic knowledge in passage retrieval for question answering</Title>
  <Section position="7" start_page="945" end_page="945" type="concl">
    <SectionTitle>
6 Conclusions
</SectionTitle>
    <Paragraph position="0"> In this paper we describe an approach for integrating linguistic information derived from dependency analyses in passage retrieval for question answering. Our retrieval component uses a multi-layer index containing various combinations of linguistic features and syntactic units extracted from a fully analyzed corpus of unrestricted Dutch text. Natural language questions are parsed in the same way.</Paragraph>
    <Paragraph position="1"> Their analyses are used to build complex queries to our extended index. We demonstrated a genetic algorithm for optimizing query parameters to improve the retrieval performance. The system was trained on questions from the CLEF competition on open-domain question answering for Dutch which are annotated with corresponding answers in the corpus.</Paragraph>
    <Paragraph position="2"> We could show a significant improvement of about 15% in mean total reciprocal rank using extended queries with optimized parameters compared with the base-line of traditional information retrieval using plain text keywords.</Paragraph>
  </Section>
class="xml-element"></Paper>