<?xml version="1.0" standalone="yes"?>
<Paper uid="P06-1133">
  <Title>Are These Documents Written from Different Perspectives? A Test of Different Perspectives Based On Statistical Distribution Divergence</Title>
  <Section position="3" start_page="0" end_page="1057" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Conflicts arise when two groups of people take very different perspectives on political, socioeconomical, or cultural issues. For example, here are the answers that two presidential candidates, John Kerry and George Bush, gave during the third presidential debate in 2004 in response to a question on abortion:  (1) Kerry: What is an article of faith for me is not something that I can legislate on somebody who doesn't share that article of faith. I believe that choice is a woman's choice. It's between a woman, God and her doctor. And that's why I support that.</Paragraph>
    <Paragraph position="1"> (2) Bush: I believe the ideal world is one in  which every child is protected in law and welcomed to life. I understand there's great differences on this issue of abortion, but I believe reasonable people can come together and put good law in place that will help reduce the number of abortions.</Paragraph>
    <Paragraph position="2"> After reading the above transcripts some readers may conclude that one takes a &amp;quot;pro-choice&amp;quot; perspective while the other takes a &amp;quot;pro-life&amp;quot; perspective, the two dominant perspectives in the abortion controversy.</Paragraph>
    <Paragraph position="3"> Perspectives, however, are not always manifested when two pieces of text together are put together. For example, the following two sentences are from Reuters newswire:  (3) Gold output in the northeast China province of Heilongjiang rose 22.7 pct in 1986 from 1985's level, the New China News Agency said.</Paragraph>
    <Paragraph position="4"> (4) Exco Chairman Richard Lacy told Reuters  the acquisition was being made from Bank of New York Co Inc, which currently holds a 50.1 pct, and from RMJ partners who hold the remainder.</Paragraph>
    <Paragraph position="5"> A reader would not from this pair of examples perceive as strongly contrasting perspectives as the Kerry-Bush answers. Instead, as the Reuters annotators did, one would label Example 3 as &amp;quot;gold&amp;quot; and Example 4 as &amp;quot;acquisition&amp;quot;, that is, as two topics instead of two perspectives.</Paragraph>
    <Paragraph position="6"> Why does the contrast between Example 1 and Example 2 convey different perspectives, but the contrast between Example 3 and Example 4 result in different topics? How can we define the impalpable &amp;quot;different perspectives&amp;quot; anyway? The definition of &amp;quot;perspective&amp;quot; in the dictionary is &amp;quot;subjective evaluation of relative significance,&amp;quot;1 but can we have a computable definition to test the existence of different perspectives?  guage, 4th ed. We are interested in identifying &amp;quot;ideological perspectives&amp;quot; (Verdonk, 2002), not first-person or secondperson &amp;quot;perspective&amp;quot; in narrative.  The research question about the definition of different perspectives is not only scientifically intriguing, it also enables us to develop important natural language processing applications. Such a computational definition can be used to detect the emergence of contrasting perspectives. Media and political analysts regularly monitor broadcast news, magazines, newspapers, and blogs to see if there are public opinion splitting. The huge number of documents, however, make the task extremely daunting. Therefore an automated test of different perspectives will be very valuable to information analysts.</Paragraph>
    <Paragraph position="7"> We first review the relevant work in Section 2. We take a model-based approach to develop a computational definition of different perspectives. We first develop statistical models for the two document collections, A and B, and then measure the degree of contrast by calculating the &amp;quot;distance&amp;quot; between A and B. How document collections are statistically modeled and how distribution difference is estimated are described in Section 3. The document corpora are described in Section 4. In Section 5, we evaluate how effective the proposed test of difference perspectives based on statistical distribution. The experimental results show that the distribution divergence can successfully separate document collections of different perspectives from other kinds of collection pairs. We also investigate if the pattern of distribution difference is due to personal writing or speaking styles.</Paragraph>
  </Section>
class="xml-element"></Paper>