<?xml version="1.0" standalone="yes"?>
<Paper uid="A00-1041">
  <Title>Answer Extraction</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> In this paper, we describe and evaluate a question-answering system based on passage retrieval and entity-extraction technology.</Paragraph>
    <Paragraph position="1"> There has long been a concensus in the Information Retrieval (IR) community that natural language processing has little to offer for retrieval systems.</Paragraph>
    <Paragraph position="2"> Plausibly, this is creditable to the preeminence of ad hoc document retrieval as the task of interest in IR.</Paragraph>
    <Paragraph position="3"> However, there is a growing recognition of the limitations of ad hoc retrieval, both in the sense that current systems have reached the limit of achievable performance, and in the sense that users' information needs are often not well characterized by document retrieval.</Paragraph>
    <Paragraph position="4"> In many cases, a user has a question with a specific answer, such as What city is it where the European Parliament meets? or Who discovered Pluto? In such cases, ranked answers with links to supporting documentation are much more useful than the ranked list of documents that standard retrieval engines produce.</Paragraph>
    <Paragraph position="5"> The ability to answer specific questions also provides a foundation for addressing quantitative inquiries such as How many times has the Fed raised interest rates this year? which can be interpreted as the cardinality of the set of answers to a specific question that happens to have multiple correct answers, like On what date did the Fed raise interest rates this year? We describe a system that extracts specific answers from a document collection. The system's performance was evaluated in the question-answering track that has been introduced this year at the TREC information-retrieval conference. The major points of interest are the following.</Paragraph>
    <Paragraph position="6"> * Comparison of the system's performance to a system that uses the same passage retrieval component, but no natural language processing, shows that NLP provides significant performance improvements on the question-answering task.</Paragraph>
    <Paragraph position="7"> * The system is designed to build on the strengths of both IR and NLP technologies. This makes for much more robustness than a pure NLP system would have, while affording much greater precision than a pure IR system would have.</Paragraph>
    <Paragraph position="8"> * The task is broken into subtasks that admit of independent development and evaluation. Passage retrieval and entity extraction are both recognized independent tasks. Other subtasks are entity classification and query classification-both being classification tasks that use features obtained by parsing--and entity ranking.</Paragraph>
    <Paragraph position="9"> In the following section, we describe the question-answering system, and in section 3, we quantify its performance and give an error analysis.</Paragraph>
  </Section>
class="xml-element"></Paper>