<?xml version="1.0" standalone="yes"?>
<Paper uid="H01-1069">
  <Title>Toward Semantics-Based Answer Pinpointing</Title>
  <Section position="3" start_page="0" end_page="0" type="intro">
    <SectionTitle>
2. WEBCLOPEDIA
</SectionTitle>
    <Paragraph position="0"> Webclopedia's architecture (Figure 1) follows the pattern outlined above: Question parsing: Using BBN's IdentiFinder [1], our parser CONTEX (Section 4) produces a syntactic-semantic analysis of the question and determines the QA type (Section 3).</Paragraph>
    <Paragraph position="1"> Query formation: Single- and multi-word units (content words) are extracted from the analysis, and WordNet synsets are used for query expansion. A Boolean query is formed. See [9]. IR: The IR engine MG [12] returns the top-ranked 1000 documents.</Paragraph>
    <Paragraph position="2"> Segmentation: To decrease the amount of text to be processed, the documents are broken into semantically coherent segments. Two text segmenter--TexTiling [5] and C99 [2]--were tried; the first is used; see [9].</Paragraph>
    <Paragraph position="3"> Ranking segments: For each segment, each sentence is scored using a formula that rewards word and phrase overlap with the question and its expanded query words. Segments are ranked. See [9] Parsing segments: CONTEX parses each sentence of the top-ranked 100 segments (Section 4).</Paragraph>
    <Paragraph position="4"> Pinpointing: For each sentence, three steps of matching are performed (Section 5); two compare the analyses of the question and the sentence; the third uses the window method to compute a goodness score.</Paragraph>
    <Paragraph position="5"> Ranking of answers: The candidate answers' scores are compared and the winner(s) are output.</Paragraph>
  </Section>
class="xml-element"></Paper>