<?xml version="1.0" standalone="yes"?>
<Paper uid="N06-1007">
  <Title>Acquisition of Verb Entailment from Text</Title>
  <Section position="8" start_page="54" end_page="55" type="concl">
    <SectionTitle>
7 Conclusion
</SectionTitle>
    <Paragraph position="0"> In this paper we proposed a novel method for automatic discovery of verb entailment relations from text, a problem that is of potential benefit for many NLP applications. The central assumption behind the method is that verb entailment relations manifest themselves in the regular co-occurrence of two verbs inside locally coherent text. Our evaluation has shown that this assumption provides a promising approach for discovery of verb entailment. The method achieves good performance, demonstrating a closer approximation to the human performance than inference rules, constructed on the basis of distributional similarity between paths in parse trees.</Paragraph>
    <Paragraph position="1"> A promising direction along which this work  can be extended is the augmentation of the current algorithm with techniques for coreference resolution. Coreference, nominal and pronominal, is an important aspect of the linguistic realization of local discourse structure, which our model did not take into account. As the experimental evaluation suggests, many verbs related by entailment occur close to one another in the text. It is very likely that many common event participants appearing in such proximity are referred to by coreferential expressions, and therefore noticeable improvement can be expected from applying coreference resolution to the corpus prior to learning entailment patterns from it.</Paragraph>
  </Section>
class="xml-element"></Paper>