<?xml version="1.0" standalone="yes"?>
<Paper uid="H05-1040">
  <Title>Proceedings of Human Language Technology Conference and Conference on Empirical Methods in Natural Language Processing (HLT/EMNLP), pages 315-322, Vancouver, October 2005. c(c)2005 Association for Computational Linguistics Enhanced Answer Type Inference from Questions using Sequential Models</Title>
  <Section position="5" start_page="321" end_page="321" type="concl">
    <SectionTitle>
5 Conclusion
</SectionTitle>
    <Paragraph position="0"> We presented a new approach to inferring the type of the answer sought by a well-formed natural language question. We introduced the notion of a span of informer tokens and extract it using a sequential graphical model with a novel feature representation derived from the parse tree of the question. Our approach beats the accuracy of recent algorithms, even ones that used max-margin methods with sophisticated kernels defined on parse trees.</Paragraph>
    <Paragraph position="1"> An intriguing feature of our approach is that when an informer (actor) is narrower than the question class (person), we can exploit direct hypernymy connections like actor to Tom Hanks, if available. Existing knowledge bases like WordNet and Wikipedia, combined with intense recent work (Etzioni et al., 2004) on bootstrapping is-a hierarchies, can thus lead to potentially large benefits.</Paragraph>
    <Paragraph position="2"> Acknowledgments: Thanks to Sunita Sarawagi for help with CRFs, and the reviewers for improving the presentation.</Paragraph>
  </Section>
class="xml-element"></Paper>