<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-1667">
  <Title>Unsupervised Relation Disambiguation with Order Identification Capabilities</Title>
  <Section position="6" start_page="574" end_page="574" type="concl">
    <SectionTitle>
4 Conclusion and Future work
</SectionTitle>
    <Paragraph position="0"> In this paper, we approach unsupervised relation disambiguation problem by using spectral-based clustering technique with diverse lexical and syntactic features derived from context. The advantage of our method is that it doesn't need any manually labeled relation instances, and pre-definition the number of the context clusters. Experiment results on the ACE corpus show that our method achieves better performance than other unsupervised methods.</Paragraph>
    <Paragraph position="1"> Currently we combine various lexical and syntactic features to construct context vectors for clustering. In the future we will further explore other semantic information to assist the relation extraction problem. Moreover, instead of cosine similarity measure to calculate the distance between context vectors, we will try other distributional similarity measures to see whether the performance of relation extraction can be improved. In addition, if we can find an effective unsupervised way to filter out unrelated entity pairs in advance, it would make our proposed method more practical.</Paragraph>
  </Section>
class="xml-element"></Paper>