<?xml version="1.0" standalone="yes"?>
<Paper uid="P98-2230">
  <Title>Machine Translation with a Stochastic Grammatical Channel Dekai Wu and Hongsing WONG HKUST</Title>
  <Section position="9" start_page="1411" end_page="1411" type="evalu">
    <SectionTitle>
6 Results
</SectionTitle>
    <Paragraph position="0"> The grammatical channel was tested in the SILC translation system. The translation lexicon was partly constructed by training on government transcripts from the HKUST English-Chinese Parallel Bilingual Corpus, and partly entered by hand.</Paragraph>
    <Paragraph position="1"> The corpus was sentence-aligned statistically (Wu, 1994); Chinese words and collocations were extracted (Fung and Wu, 1994; Wu and Fung, 1994); then translation pairs were learned via an EM procedure (Wu and Xia, 1995). Together with hand-constructed entries, the resulting English vocabulary is approximately 9,500 words and the Chinese vocabulary is approximately 14,500 words, with a many-to-many translation mapping averaging 2.56 Chinese translations per English word. Since the lexicon's content is mixed, we approximate translation probabilities by using the unigram distribution of the target vocabulary from a small monolingual corpus. Noise still exists in the lexicon.</Paragraph>
    <Paragraph position="2"> The Chinese grammar we used is not tight-it was written for robust parsing purposes, and as such it over-generates. Because of this we have not yet been able to conduct a fair quantitative assessment of objective 3. Our productions were constructed with reference to a standard grammar (Beijing Language and Culture Univ., 1996) and totalled 316 productions. Not all the original productions are mirrored, since some (128) are unary productions, and others are Chinese-specific lexical constructions like S ~ ~-~ S NP ~ S, which are obviously unnecessary to handle English. About 27.7% of the non-unary Chinese productions were mirrored and the total number of productions in the final ITG is 368.</Paragraph>
    <Paragraph position="3"> For the experiment, 222 English sentences with a maximum length of 20 words from the parallel corpus were randomly selected. Some examples of the output are shown in Figure 2. No morphological processing has been used to correct the output, and up to now we have only been testing with a bigram model trained on extremely small corpus.</Paragraph>
    <Paragraph position="4"> With respect to objective 1 (increasing translation speed), the new model is very encouraging. Table 1 shows that over 90% of the samples can be processed within one minute by the grammatical channel model, whereas that for the SBTG channel model is about 50%. This demonstrates the stronger  constraints on the search space given by the SITG.</Paragraph>
    <Paragraph position="5"> The natural trade-off is that constraining the structure of the input decreases robustness somewhat. Approximately 13% of the test corpus could not be parsed in the grammatical channel model.</Paragraph>
    <Paragraph position="6"> As mentioned earlier, this figure is likely to vary widely depending on the characteristics of the target grammar. Of course, one can simply back off to the SBTG model when the grammatical channel rejects an input sentence.</Paragraph>
    <Paragraph position="7"> With respect to objective 2 (improving meaning-preservation accuracy), the new model is also promising. Table 2 shows that the percentage of meaningfully translated sentences rises from 26% to 32% (ignoring the rejected cases). 7 We have judged only whether the correct meaning is conveyed by the translation, paying particular attention to word order and grammaticality, but otherwise ignoring morphological and function word choices.</Paragraph>
  </Section>
class="xml-element"></Paper>