<?xml version="1.0" standalone="yes"?>
<Paper uid="P03-2005">
  <Title>An Adaptive Approach to Collecting Multimodal Input</Title>
  <Section position="8" start_page="0" end_page="0" type="evalu">
    <SectionTitle>
7 Results
</SectionTitle>
    <Paragraph position="0"> The enhanced module was tested using the data collected in previous tests and further online tests.</Paragraph>
    <Paragraph position="1"> The average delay in determining the end of turn reduced to 1.3 secs. This represents a 40% improvement on the earlier results. Also based on online experiments, with the same users and tasks, the number of times users repeated their input was reduced to 2% and collection errors reduced to 3% (compared to 8% and 6% respectively). The improvement was partly due to the reduced delay in the determination of the end of the user's turn and also due to prediction of the preferred interaction style. It was also observed that the performance increased by a further 5% by using online learning.</Paragraph>
    <Paragraph position="2"> The results demonstrate the effectiveness of the proposed approach to the robustness and temporal performance of MMIF.</Paragraph>
  </Section>
class="xml-element"></Paper>