<?xml version="1.0" standalone="yes"?>
<Paper uid="N04-1029">
  <Title>Comparison of Two Interactive Search Refinement Techniques</Title>
  <Section position="7" start_page="0" end_page="0" type="concl">
    <SectionTitle>
5 Conclusions and future work
</SectionTitle>
    <Paragraph position="0"> In this paper we presented two user-assisted search refinement techniques: (1) inviting the user to select from the clarification form a number of sentences that may represent relevant documents, and then using the documents whose sentences were selected for query expansion.</Paragraph>
    <Paragraph position="1"> (2) showing to the user a list of noun phrases, extracted  5 rel. documents 6 rel. documents 7 rel. documents 15 rel. documents from the initial document set, and then expanding the query with the terms from the user-selected phrases.</Paragraph>
    <Paragraph position="2"> The evaluation results suggest that the second expansion method overall is more promising than the first, demonstrating statistically significant performance improvement over the baseline run. More analysis needs to be done to determine the key factors influencing the performance of both methods.</Paragraph>
    <Paragraph position="3"> The focus of our experiments in the HARD track of TREC-12 was on developing effective methods of gathering and utilising the user's relevance feedback. Another major goal of the HARD track, which we did not address this time, is to promote research into how contextual and extra-linguistic information about the user and the user's search task could be harnessed to achieve high accuracy retrieval. To effectively use information such as user's familiarity with the topic, the purpose of the user's search or the user's genre preferences we need more complex linguistic and stylistic analysis techniques. We plan to address these issues in the next year's entry.</Paragraph>
  </Section>
class="xml-element"></Paper>