I am the outbound campaign manager and I am currently trying to optimize our predictive dialer performance. We’ve implemented ‘Speech Analytics’ topics to flag calls where customers mention our main competitor. However, my ‘Competitor Mention’ report is missing about fifty percent of the actual occurrences. When I listen to the recordings, I can hear the customer say the competitor’s name, but the agent is often speaking over them at the exact same moment. Does the speech analytics engine struggle with ‘Over-talk’, and is there a way to prioritize the customer’s audio channel for topic detection in a dual-channel recording?
I am completely fed up with the topic detection accuracy! I am the speech analytics manager and I’ve been telling the management for months that ‘Over-talk’ is the number one killer of our ROI! The engine attempts to transcribe both channels, but if the audio is mixed or if there is significant crosstalk, the phonetic analyzer gets completely confused. It doesn’t matter how dual-channel your recording is if the transcription engine can’t separate the speakers during a heated disagreement! I’m tired of seeing these ‘Missed Topics’ in our reports just because our agents won’t let the customers finish their sentences!
As an Edge appliance admin, I can tell you that the ‘Over-talk’ issue is often made worse by the ‘Echo Cancellation’ settings on the agent’s headset. If the agent’s headset is leaking audio back into the microphone, it creates a ‘Double-Talk’ scenario that the speech analytics engine cannot resolve. You should ensure that all your agents are using Genesys-certified noise-canceling headsets with ‘Side-tone’ disabled. It sounds like a hardware fix for a software problem, but cleaning up the raw audio input is the only way to improve the transcription accuracy for those competitor mentions.
Hello! I am the routing optimization engineer and I’ve seen how these speech analytics gaps can impact our bullseye routing logic! While hardware is important, you should also look into the ‘Speaker Diarization’ settings in your Speech Analytics configuration. You can actually set the engine to ‘Prefer External Participant’ for specific topic searches. By telling the AI to prioritize the customer’s phonetic stream for competitor keywords, you can significantly reduce the ‘Masking’ effect caused by the agent’s voice. It’s such a brilliant way to fine-tune your analytics without having to retrain your entire agent workforce on conversational etiquette!