Answering Machine Detector for Call Centers

Autores UPV
CONGRESO Answering Machine Detector for Call Centers


This paper presents a system for detecting answering machines in call centers. The task is to distinguish if the audio from the other side of the telephone line belongs to a recording or to a person. The challenge of this system is to make a decision after processing the first three seconds of audio. The detector system tells the call center whether hanging up or transferring the call to a human agent. Three sources of information are used for making the decision: the time to pick up the phone, the percentage of silence in the three seconds of audio and the recognized word sequence. The system was tested in a real scenario but few calls were used.