Customer service quality inspection method, system and device based on voice separation and medium

Abstract

The invention relates to a customer service quality inspection method, system and device based on voice separation and a medium. The method comprises the steps of collecting a call record of customerservice staff and a customer; carrying out human voice separation on the call record by adopting a pre-trained noise reduction network model to obtain a first audio file of the customer service staffin the call record and a maximum matching degree; and if the matching degree is higher than a preset threshold, performing quality inspection on the service of the customer service staff correspondingto the maximum matching degree according to the first audio file. According to the invention, the human voice separation can be carried out on the call record, only the voice of the customer servicestaff needing quality inspection is reserved, the voice of customers not needing quality inspection is eliminated, the matching degree of the customer service staff is given so as to further confirm the identity of the customer service staff, and the workload of the quality inspection staff is greatly reduced.

Description

technical field [0001] The invention relates to the field of voice recognition, in particular to a customer service quality inspection method, system, equipment and medium based on human voice separation. Background technique [0002] In a call scenario, at least two people participate in the conversation. In some business scenarios, there is a need to distinguish different speakers and identify their speech content, and further perform speech analysis. However, due to technical limitations, the speech to be recognized and separated Sometimes it will be saved in the same channel or the voices of multiple people will be received at the same time in a noisy environment, which will make it difficult to further identify the voice of the target speaker from the mixed voice of multiple speakers and perform speech analysis. This is the so-called "cocktail party problem". [0003] At present, in sales offices and other telemarketing scenarios, in order to ensure the accuracy and pr...

AI Technical Summary

Problems solved by technology

[0002] In a call scenario, at least two people participate in the conversation. In some business scenarios, there is a need to distinguish different speakers and identify their speech content, and further perform speech analysis. However, due to technical limitations, the speech to be recognized and separated Sometimes it will be saved in the same channel or the voices of multiple people will be received at the same time in a noisy environment, which will make it difficult to further identify the voice of the target speaker from the mixed voice of multiple speakers and perform speech analysis. It's called the "cocktail party

Images