<?xml version="1.0" standalone="yes"?>
<Paper uid="W03-1717">
  <Title>Learning Verb-Noun Relations to Improve Parsing</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Computer analysis of natural language sentences is a challenging task largely because of the ambiguities in natural language syntax. In Chinese, the lack of inflectional morphology often makes the resolution of those ambiguities even more difficult.</Paragraph>
    <Paragraph position="1"> One type of ambiguity is found in the verb-noun sequence which can appear in at least two different relations, the verb-object relation and the modifier-head relation, as illustrated in the following phrases.</Paragraph>
    <Paragraph position="2">  (1) Deng Ji Shou Xu De Fei Yong  dengji shouxu de feiyong register procedure DE expense &amp;quot;the expense of the registration procedure&amp;quot; (2) Ban Li Shou Xu De Fei Yong banli shouxu de feiyong handle procedure DE expense &amp;quot;the expense of going through the procedure&amp;quot; In (1), the verb-noun sequence &amp;quot;Deng Ji Shou Xu &amp;quot; is an example of the modifier-head relation while &amp;quot;Ban Li Shou Xu &amp;quot; in (2) is an example of the verb-object relation. The correct analyses of these two phrases are given in Figure 1 and Figure 2, where &amp;quot;RELCL&amp;quot; stands for &amp;quot;relative clause&amp;quot;:  However, with the set of grammar rules that cover the above phrases and without any semantic or collocational knowledge of the words involved, there is nothing to prevent us from the wrong analyses in Figure 3 and Figure 4.</Paragraph>
    <Paragraph position="3">  To rule out these wrong parses, we need to know that Deng Ji is a typical modifier ofShou Xu while Ban Li typically takes Shou Xu as an object. The question is how to acquire such knowledge automatically. In the rest of this paper, we will present a learning procedure that learns those relations by processing a large corpus with a chart-filter, a tree-filter and an LLR filter. The approach is in the spirit of Smadja (1993) on retrieving collocations from text corpora, but is more integrated with parsing. We will show in the evaluation section how much the learned knowledge can help improve sentence analysis.</Paragraph>
  </Section>
class="xml-element"></Paper>