<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-3255">
  <Title>Efficient Decoding for Statistical Machine Translation with a Fully Expanded WFST Model</Title>
  <Section position="6" start_page="0" end_page="0" type="evalu">
    <SectionTitle>
5 Experiments
</SectionTitle>
    <Paragraph position="0"/>
    <Section position="1" start_page="0" end_page="0" type="sub_section">
      <SectionTitle>
5.1 Effect of Full Expansion
</SectionTitle>
      <Paragraph position="0"> To clarify the effectiveness of a full-expansion approach, we compared the computational costs while using the same decoder with both dynamic composition and static composition, a full-expansion model in other words. In the forward beam-search, any hypothesis whose probability is lower than a48a100a99a100a48a100a101 of the top of the hypothesis list is pruned. In this experiment, permutation is restricted, and words can be moved 6 positions at most. The translation model was trained by GIZA++ (Och and Ney, 2003), and the trigram was trained by the CMU-Cambridge Statistical Language Modeling Toolkit v2 (Clarkson and Rosenfeld, 1997).</Paragraph>
      <Paragraph position="1"> For the experiment, we used a Japanese-to-English bilingual corpus consisting of example sentences for a rule-based machine translation system. Each language sentence is aligned in the corpus. The total number of sentence pairs is 20,204.</Paragraph>
      <Paragraph position="2"> We used 17,678 pairs for training and 2,526 pairs for the test. The average length of Japanese sentences was 8.4 words, and that of English sentences was 6.7 words. The Japanese vocabulary consisted of 15,510 words, and the English vocabulary was 11,806 words. Table 1 shows the size of the WFSTs used in the experiment. In these WFSTs, special symbols that express beginning and end of sentence are added to the WFSTs described in the previous section. The NIST score (Doddington, 2002) and BLEU Score (Papineni et al., 2002) were used to measure translation accuracy.</Paragraph>
      <Paragraph position="3"> Table 2 shows the experimental results. The full-expansion model provided translations more than 10 times faster than conventional dynamic composition submodels without degrading accuracy. However, the NIST scores are slightly different. In the course of composition, some paths that do not reach the final states are produced. In the full-expansion model these paths are trimmed. These trimmed paths may cause a slight difference in NIST scores.</Paragraph>
    </Section>
    <Section position="2" start_page="0" end_page="0" type="sub_section">
      <SectionTitle>
5.2 Effect of Ambiguity Reduction
</SectionTitle>
      <Paragraph position="0"> To show the effect of ambiguity reduction, we compared the translation results of three different models. Model a102 is the full-expansion model described above. Model a81 is a reduced model by using our proposed method with a 0.9 a34 a43 threshold. Model a81a82a103 is a reduced model with the statistics of the decoder without using the correct translation WFST. In other words, a81a82a103 reduces the states of the full-expansion model more roughly than a81 . The a34  threshold for a81a82a103 is set to 0.85 so that the size of the produced WFST is almost the same as a81 . Table 3 shows the model size. To obtain decoder statistics for calculating a34  , all of the sentence pairs in the training set were used. When obtaining the statistics, any hypothesis whose probability is lower than a48a100a99a100a48a100a101 a36a31a104a105 of the top of the hypothesis list is pruned in the forward beam-search.</Paragraph>
      <Paragraph position="1"> The translation experiment was conducted by successively changing the beam width of the forward search. Figures 7 and 8 show the results of the translation experiments, revealing that our proposed model can reduce the decoding time by approximately half. This model can reduce decoding time to a much greater extent than the rough reduction model, indicating that our state merging criteria are valid.</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>