ISCA Archive Interspeech 2021
ISCA Archive Interspeech 2021

Audio Segmentation Based Conversational Silence Detection for Contact Center Calls

Krishnachaitanya Gogineni, Tarun Reddy Yadama, Jithendra Vepa

In a typical contact-center call, more than 35% of the call has neither the contact-center agent nor the customer speaking, we usually refer to such areas in the call as Conversational Silences. Conversational silences comprise mostly of hold-music, automatic-recorded-messages, or just silences when the agent or customer is engaged in some off-call work. Most of these conversational silences negatively affect important KPIs for call-centers, like dead-airs affect customer satisfaction, long-holds affect average call handling time and so on. In this paper we showcase how Observe.AI helps contact-centers identify agents who are breaching accepted levels of conversational silences by using an in-house Audio Segmenter system paired with an NLP system to classify the contexts around these Conversational Silences. This solution is provided by Observe.AI to hundreds of contact centers who use it to improve their average call handling time and customer satisfaction scores.


Cite as: Gogineni, K., Yadama, T.R., Vepa, J. (2021) Audio Segmentation Based Conversational Silence Detection for Contact Center Calls. Proc. Interspeech 2021, 2349-2350

@inproceedings{gogineni21_interspeech,
  author={Krishnachaitanya Gogineni and Tarun Reddy Yadama and Jithendra Vepa},
  title={{Audio Segmentation Based Conversational Silence Detection for Contact Center Calls}},
  year=2021,
  booktitle={Proc. Interspeech 2021},
  pages={2349--2350}
}