<?xml version="1.0" standalone="yes"?>
<Paper uid="P06-1093">
  <Title>Automatic Generation of Domain Models for Call Centers from Noisy Transcriptions</Title>
  <Section position="8" start_page="743" end_page="743" type="concl">
    <SectionTitle>
6 Discussion and Future Work
</SectionTitle>
    <Paragraph position="0"> We have shown that it is possible to retrieve useful information from noisy transcriptions of call center voice conversations. We have shown that the extracted information can be put in the form of a model that succinctly captures the domain and provides a comprehensive view of it. We briefly showed through experiments that this model is an accurate description of the domain. We have also suggested useful scenarios where the model can be used to aid and improve call center performance.</Paragraph>
    <Paragraph position="1"> A call center handles several hundred-thousand calls per year in various domains. It is very difficult to monitor the performance based on manual processing of the calls. The framework presented in this paper, allows a large part of this work to be automated. A domain specific model that is automatically learnt and updated based on the voice conversations allows the call center to identify problem areas quickly and allocate resources more effectively.</Paragraph>
    <Paragraph position="2"> In future we would like to semantically cluster the topic specific information so that redundant topics are eliminated from the list. We can use Automatic Taxonomy Generation(ATG) algorithms for document summarization (Kummamuru et al., 2004) to build topic taxonomies. We would also like to link our model to technical manuals, catalogs, etc. already available on the different topics in the given domain.</Paragraph>
    <Paragraph position="3"> Acknowledgements: We thank our colleagues Raghuram Krishnapuram and Sreeram Balakrishnan for helpful discussions. We also thank Olivier Siohan from the IBM T. J. Watson Research Center for providing us with call transcriptions.</Paragraph>
  </Section>
class="xml-element"></Paper>