<?xml version="1.0" standalone="yes"?>
<Paper uid="W97-0402">
  <Title>A Dialogue Analysis Model with Statistical Speech Act Processing for Dialogue Machine Translation*</Title>
  <Section position="7" start_page="13" end_page="14" type="evalu">
    <SectionTitle>
5 Experiments and Results
</SectionTitle>
    <Paragraph position="0"> In order to experiment the proposed model, we used 70 dialogues recorded in real fields such as hotel reservation and airline reservation. These 70 dialogues consist of about 1,700 utterances, 8,319 words total. Each utterance in dialogues was annotated with speech acts (SA) and with discourse structure information (DS). DS is an index that represents the hierarchical structure of discourse. Table 2 shows the distribution of speech acts in this dialogue corpus. The following shows a part of an annotated dialogue corpus.</Paragraph>
    <Paragraph position="1">  We test two models in order to verify the efficiency of the proposed model. Model-I is the proposed model based on linear recency, where an utterance U/ is always connected to the previous utterance Ui-1. Model-II is the model based on hierarchical recency. Table 3 shows the average accuracy of two models.</Paragraph>
    <Paragraph position="2"> Accuracy figures shown in table 3 are computed by counting utterances that have a correct speech act and a correct discourse relation. In the closed experiments, Modelq achieved 68.48 % accuracy for the top candidate and 76.30 % for the top four candidates. In contrast, the proposed model, Model-II, achieved 78.59 % accuracy for the top candidate and 99.06 % for the top four candidates. Errors in Model-I occurred, because the hierarchical structure of dialogues was not considered. Although dialogue corpus are relatively small, the experimental results showed that the proposed model is efficient for analyzing dialogues.</Paragraph>
  </Section>
class="xml-element"></Paper>