<?xml version="1.0" standalone="yes"?>
<Paper uid="I05-3016">
  <Title>Resolving Pronominal References in Chinese with the Hobbs Algorithm</Title>
  <Section position="2" start_page="0" end_page="116" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> The goal of this study is pronoun resolution, including null/zero pronouns, in Chinese. There has been extensive research for many years into computational approaches to automatic anaphora resolution in English, and increasingly in other languages as well (Mitkov, 1999; Mitkov, 2002).</Paragraph>
    <Paragraph position="1"> Yet although there have been countless linguistic studies in Chinese on anaphora and zero anaphora (for example, (Huang, 1984; Huang, 1994; Yang et al., 1999) just to illustrate the range), the published computational work to date is limited to just a few studies (Chen, 1992; Yeh and Chen, 2001; Yeh and Chen, 2005).</Paragraph>
    <Paragraph position="2"> In 1978 Jerry Hobbs proposed an algorithm for the resolution of pronominal coreference in English (Hobbs, 1978). The performance of this algorithm has frequently been used as a baseline reference for computational methods in English.</Paragraph>
    <Paragraph position="3"> The most basic version of the Hobbs algorithm is subject biased, relying on a basic strategy of leftto-right, breadth-first searches, subject to a few structural constraints.</Paragraph>
    <Paragraph position="4"> Chinese, like English, is an SVO language.</Paragraph>
    <Paragraph position="5"> Chinese has also been regarded as a topic-comment language. From either viewpoint, it is worth examining how well the left-to-right, SVObiased process of the Hobbs algorithm works for Chinese, perhaps so it could be used as a base-line against which to measure other automated approaches to Chinese anaphora resolution.</Paragraph>
    <Paragraph position="6"> While Chinese and English are both SVO languages, they differ in another important parameter: Chinese is a pro-drop language, while standard English is not. Thus it will be of particular interest to see how well the Hobbs algorithm performs when proposing antecedents for zero pronouns. null The Hobbs algorithm operates on parsed sentences. In order to evaluate its performance on zero pronouns as well as overt ones, it would be useful to have text that already has the locations of  the zero pronouns marked. Because the Penn Chinese Treebank has overt strings to denote the positions of dropped arguments, test sentences were selected from that corpus.</Paragraph>
  </Section>
class="xml-element"></Paper>