<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-3404">
  <Title>You Are What You Say: Using Meeting Participants' Speech to Detect their Roles and Expertise</Title>
  <Section position="9" start_page="28" end_page="29" type="concl">
    <SectionTitle>
8 Conclusions and Future Work
</SectionTitle>
    <Paragraph position="0"> In this paper we have discussed our current approach to detecting the functional and expertise based roles of meeting participants. We have induced decision  trees that use simple and robust speech based features to perform the role detection. We have used a very simple evidence aggregation mechanism to arrive at a single role assignment per meeting participant over a sequence of meetings, and have shown that we can achieve up to 83% accuracy on unseen test data using this mechanism. Additionally we have shown that by aggregating evidence across a sequence of meetings, we perform better than if we were to use a single meeting to perform the role detection. As future work we plan to remove the constraints that we have currently imposed - namely, we will attempt to learn new roles in test data that do not exist in training data. Additionally, we will attempt to use this role information as inputs to downstream meeting understanding tasks such as automatic topic detection and action item detection.</Paragraph>
  </Section>
class="xml-element"></Paper>