<?xml version="1.0" standalone="yes"?>
<Paper uid="N06-4011">
  <Title>AUTOMATEDQUALITYMONITORINGFORCALLCENTERSUSINGSPEECHANDNLP TECHNOLOGIES</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
ABSTRACT
</SectionTitle>
    <Paragraph position="0"> This paper describesan automatedsystemfor assigningquality scoresto recordedcall center conversations. The systemcombinesspeechrecognition,patternmatching,andmaximumentropy null classificationto rank calls according to their measured quality.</Paragraph>
    <Paragraph position="1"> Callsat both ends of the spectrumare flaggedas &amp;quot;interesting&amp;quot;and made availablefor furtherhuman monitoring.In this process,the ASR transcriptis used to answer a set of standardqualitycontrol questionssuchas &amp;quot;didthe agentuse courteouswordsand phrases,&amp;quot; and to generatea question-basedscore. This is interpolatedwith the probabilityof a call being &amp;quot;bad,&amp;quot; as determinedby maximum entropy operatingon a set of ASR-derived featuressuch as &amp;quot;maximumsilencelength&amp;quot;and the occurrenceof selectedn-gramword sequences. The systemis trainedon a set of calls with associated manual evaluationforms. We present precisionand recall results fromIBM's NorthAmericanHelpDeskindicatingthat for a given amount of listeningeffort, this system triples the number of bad calls that are identified,over the current policy of randomlysamplingcalls. The applicationthat will be demonstratedis a research prototypethat was built in conjunctionwith IBM's North American call centers.</Paragraph>
  </Section>
class="xml-element"></Paper>