<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-2904">
  <Title>Improved Large Margin Dependency Parsing via Local Constraints and Laplacian Regularization</Title>
  <Section position="10" start_page="26" end_page="27" type="evalu">
    <SectionTitle>
8 Experimental Results
</SectionTitle>
    <Paragraph position="0"> We tested our method experimentally on the Chinese  in CTB are constituency structures. We converted them into dependency trees using the same method and head- nding rules as in (Bikel, 2004). Following (Bikel, 2004), we used Sections 1-270 for training, Sections 271-300 for testing and Sections 301325 for development. We experimented with two sets of data: CTB-10 and CTB-15, which contains sentences with no more than 10 and 15 words respectively. Table 1, Table 2 and Table 3 show our experimental results trained and evaluated on Chinese Treebank sentences of length no more than 10, using the standard split. For any unseen link in the new sentences, the weight is computed as the similarity weighted average of similar links seen in the training corpus. The regularization parameter a116 was set by 5-fold cross-validation on the training set.</Paragraph>
    <Paragraph position="1"> We evaluate parsing accuracy by comparing the undirected dependency links in the parser outputs against the undirected links in the treebank. We dene the accuracy of the parser to be the percentage of correct dependency links among the total set of dependency links created by the parser.</Paragraph>
    <Paragraph position="2"> Table 1 and Table 2 show that training based on the more re ned local loss is far superior to training with the global loss of standard large margin training, on both the test and development sets. Parsing accuracy also appears to increase with the introduction of each new feature. Notably, the pointwise mutual information and distance features signi cantly improve parsing accuracy and yet we know of no other research that has investigated these features in this context. Finally, we note that Laplacian regularization improved performance as expected, but not for the global loss, where it appears to systematically degrade performance (n/a results did not complete in time). It seems that the global loss model may have been over-regularized (Table 3). However, we have picked the a116 parameter which gave us the  best resutls in our experiments. One possible explanation for this phenomenon is that the interaction between the Laplician regularization in training and the similarity smoothing in parsing, since distributional word similarities are used in both cases. Finally, we compared our results to the probabilistic parsing approach of (Wang et al., 2005), which on this data obtained accuracies of 0.7631 on the CTB test set and 0.6104 on the development set. However, we are using a much simpler feature set here.</Paragraph>
  </Section>
class="xml-element"></Paper>