摘要 |
<P>PROBLEM TO BE SOLVED: To discriminate a speaker of recorded data easily regarding a speaker discrimination program recording a telephone call, a speaker discrimination device, and a speaker discrimination method. <P>SOLUTION: A voice packet from a calling message issuance terminal is recorded as an L channel, and a voice packet from a non-issuance terminal is recoded as an R channel. Access to call information files Ca and Cb is performed in retrieval, and whether or not a GW address of a designated speaker is an L-side GW address or an R-side GW address is determined. Since a GW address of a service provider is the L-side GW address in the call information file Ca, it is found that a voice of the service provider has been recorded by the L channel. Accordingly, a word coincident with a retrieval keyword is extracted from among words of the L channel within an intermediate file Ta. Thus, since keyword retrieval is performed per speaker, the efficiency of a confirmation operation of telephone call contents is improved regardless of whether either a service provider or a customer is a calling party or a called party. <P>COPYRIGHT: (C)2010,JPO&INPIT |