<?xml version="1.0" standalone="yes"?>
<Paper uid="W02-1024">
  <Title>A Hybrid Approach to Natural Language Web Search</Title>
  <Section position="8" start_page="0" end_page="0" type="concl">
    <SectionTitle>
8 Conclusions and Future Work
</SectionTitle>
    <Paragraph position="0"> In this paper, we described and evaluated RISQUE, a hybrid system for performing natural language search on a large company public website. RISQUE utilizes a two-pronged approach to generate hit lists for answering natural language questions. On the one hand, RISQUE employs a hub-page identifier to retrieve, when possible, a hub-page for the most salient NP in the question. On the other hand, RISQUE adopts a statistical iterative query formulation and retrieval mechanism that generates new queries by applying transformation rules to previously-issued queries. By employing these two components in parallel, RISQUE takes advantages of both knowledge-driven and machine learning approaches, and achieves an overall 137% relative improvement in the number of questions correctly answered on an unseen test set, compared to a baseline of 2NP keyword queries.</Paragraph>
    <Paragraph position="1"> In our current work, we are focusing on expanding system coverage to other domains. In particular, we plan to investigate semi-automatic methods for extracting ontological knowledge from existing webpages and databases.</Paragraph>
  </Section>
class="xml-element"></Paper>