<?xml version="1.0" standalone="yes"?>
<Paper uid="C02-1056">
  <Title>Paraphrasing of Chinese Utterances</Title>
  <Section position="6" start_page="3" end_page="3" type="evalu">
    <SectionTitle>
5 Experiment and Evaluation
</SectionTitle>
    <Paragraph position="0"> A paraphrasing experiment was carried out on the paraphrase corpus using the implemented paraphraser and the obtained patterns. In order to get the same effect as that of using open test data, each pattern was not applied to the sentence from which the pattern was generalized. For 45,110 test sentences, 4,908 test sentences (about 10.9%) were paraphrased. From the 4,908 test sentences, 8,183 paraphrases were generated and the average number of paraphrases for one test sentence was 1.66. The generated paraphrases were evaluated by Chinese natives from two viewpoints, i.e., naturalness and meaning-retaining, with their corresponding test sentences. As a result, 7,226 generated paraphrases were correct and an 88% accuracy was achieved. The experimental result is shown in Table 2.</Paragraph>
    <Paragraph position="1">  Three examples of the paraphrasing results are given below.</Paragraph>
    <Paragraph position="2">  (As for room, please give me one with a nice view.) In the Input of 5-1 there is an expression of repeated interrogation &amp;quot;547554&amp;quot; that consists of an affirmation &amp;quot;54(can)&amp;quot; and a negation &amp;quot;7554 (can not)&amp;quot;. After the application of the patterns of deleting negative expressions, Paraphrase 1 and Paraphrase 2 were generated. Both paraphrases are in affirmative form and both are correct. From the Input of 5-2, five paraphrases were generated only by reordering words. Paraphrases 1, 2, 3, 4 and 5 are all correct. In the Input of 5-3, the order of the predicate &amp;quot;496745&amp;quot; and the object &amp;quot;65605156696462&amp;quot; is inverted. After the patterns of inserting &amp;quot;76&amp;quot; were applied, Paraphrase 1 was obtained. Then, the patterns of deleting &amp;quot;76&amp;quot; were applied to this generated paraphrase and Paraphrase 2 and Paraphrase 3 were obtained. In Paraphrase 3 the common word order was recovered. Finally, the patterns of reordering words were applied to Paraphrases 1, 2, 3 and the input. Paraphrase 4 was obtained from Paraphrase 3. Paraphrases 1, 2, 3 and 4 are all correct.</Paragraph>
    <Paragraph position="3"> From the experimental results we see that the proposed approach can realize the goal of simplifying the expressions of the inputs and increasing variations with a high level of accuracy. If one of the paraphrased results can be translated, we can say that the paraphraser is effective in the translation system.</Paragraph>
    <Paragraph position="4"> Through the analysis of wrong results, we found two reasons for paraphrasing errors. One reason is that some constituents or modification relations are incorrectly recognized based on the obtained paraphrasing patterns. For example, when pattern 3-6 was applied to the sentence &amp;quot;49674555665033696462(A room for two people, please.)&amp;quot;, the quantity phrase &amp;quot;5566(two)&amp;quot; was wrongly recognized as modifying &amp;quot;5033696462(a room where people can live)&amp;quot;, whereas it in fact modifies the noun &amp;quot;50(people)&amp;quot;. Because of this wrong recognition, the generated sentence was &amp;quot;50336964624967455566 (Please give me two rooms where people can live.)&amp;quot;. Its meaning is different from the input and therefore the result is not correct. The other reason for paraphrasing errors is that there were errors in part-of-speech tagging. For example, the word &amp;quot;67&amp;quot;in &amp;quot;49674579344143365946 (Please gift-wrap this)&amp;quot; was tagged as a verb while it really acts as a preposition. The wrong tagging resulted in the wrong application of patterns.</Paragraph>
  </Section>
class="xml-element"></Paper>