TopicListener: Observing Key Topics from Multi-channel Speech Audio Streams

Speech audio often encapsulates huge volumes of information which traditionally has been challenging to mine and analyse using automated methods. For example, call centres often handle many simultaneous telephone conversations between customers and call centre agents where, apart from relying on limited manual reporting by individual call centre agents, the content, themes and topics of the conversations are not analysed in any depth. In recent years there have been significant improvements in both the accuracy and cost of automated speech-to-text transcription technologies which can be applied in the call centre environment. We introduce TopicListener, which combines advanced topic modelling techniques with automatic speech transcription to identify key themes and topics across large volumes of recorded audio conversions as well as providing a novel means to explore and visualise the correlation and evolution of topics over time.