<?xml version="1.0" standalone="yes"?>
<Paper uid="E06-1003">
  <Title>Weakly Supervised Approaches for Ontology Population</Title>
  <Section position="8" start_page="22" end_page="23" type="concl">
    <SectionTitle>
6 Conclusions and future work
</SectionTitle>
    <Paragraph position="0"> In this paper we presented a new weakly supervised approach for Ontology Population, called Class-Example, and confronted it with two other methods. Experimental results show that the Class-Example approach has best performance. In particular, it reached 65% of accuracy, outperforming in our experimental framework the state-of-the-art Class-Word method by 42%. Moreover, for location names the method reached accuracy of 78%. Although the experiments are not comparable, we would like to state that some supervised approaches for fine-grained Named Entity classification, e.g. (Fleischman, 2001), have similar accuracy. On the other hand, the presented weakly supervised Class-Example approach requires as a training data only a list of terms for each class under consideration. Training examples can be automatically acquired from existing ontologies or other sources, since the approach imposes virtually no restrictions on them. This makes our weakly supervised methodology applicable on larger scale than supervised approaches, still having significantly better performance than the unsupervised ones.</Paragraph>
    <Paragraph position="1"> In our experimental framework we used syntactic features extracted from dependency parse trees  and we put forward a novel model for the representation of a syntactically parsed corpus. This model allows for performing a comprehensive extraction of syntactic features from a corpus including more complex second-order ones, which resulted in an improvement of performance. This and other empirical observations not described in this paper lead us to the conclusion that the performance of an Ontology Population system improves with the increase of the types of syntactic features under consideration.</Paragraph>
    <Paragraph position="2"> In our future work we consider applying our Ontology Population methodology to more semantic categories and to experiment with other types of syntactic features, as well as other types of feature-weighting formulae and learning algorithms. We consider also the integration of the approach in a Question Answering or Information Extractionsystem, whereitcanbeusedtoperform fine-grained type checking.</Paragraph>
  </Section>
class="xml-element"></Paper>