<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-2912">
  <Title>Unsupervised Parsing with U-DOP</Title>
  <Section position="6" start_page="90" end_page="90" type="concl">
    <SectionTitle>
4 Conclusions
</SectionTitle>
    <Paragraph position="0"> We have shown that the general DOP approach can be generalized to unsupervised learning, effectively leading to a single model for both supervised and unsupervised parsing. Our new model, U-DOP, uses all subtrees from (in principle) all binary trees of a set of sentences to compute the most probable parse trees for (new) sentences. Although heavy pruning of trees is necessary to make our approach feasible in practice, we obtained competitive results on English, German and Chinese data. Our parsing results are similar to the performance of a binarized supervised PCFG on the WSJ [?] 40 sentences. This triggers the provocative question as to whether it is possible to beat supervised parsing by unsupervised parsing. To cope with the problem of evaluation, we propose to test U-DOP in specific applications rather than on hand-annotated data.</Paragraph>
  </Section>
class="xml-element"></Paper>