A speech quality detection method and device

Abstract

The invention discloses a speech quality detection method and device, which relates to the technical field of speech detection, and mainly aims at improving the comprehensiveness and accuracy of speech quality detection. The main technical scheme of the invention is characterized by converting the speech to be quality-checked into the text to be quality-checked; extracting quality inspection itemsone by one from a quality inspection item collection, wherein the quality inspection items are used for detecting whether the text to be inspected conforms to requirements; using a pre-trained machine learning model to detect whether the text to be inspected contains content matching with the quality inspection item; according to the matching result and the detection rule corresponding to the quality inspection entry, determining whether the speech to be inspected is qualified or not. The invention is used for the normative detection of the voice data content.

Description

technical field [0001] The invention relates to the technical field of voice detection, in particular to a voice quality inspection method and device. Background technique [0002] Customer service voice quality inspection is to promote enterprise marketing, market development, improve customer satisfaction, improve customer service, and at the same time evaluate the work of customer service personnel. The call center generates a large number of telephone recordings every day. Through the quality inspection of these recordings, it is tested whether the customer service personnel use standardized language and whether the service is in place to meet the needs of customers. [0003] The traditional voice quality inspection work is performed manually by quality inspectors. This method not only consumes a lot of manpower when faced with a large amount of voice data, but also causes quality inspection problems due to the different quality inspection scales of inspectors. unstable...

AI Technical Summary

Problems solved by technology

[0003] The traditional voice quality inspection work is performed manually by quality inspectors. This method not only consumes a lot of manpower when faced with a large amount of voice data, but also causes quality inspection problems due to the different quality inspection scales of inspectors. unstable

In order to avoid this situation, two quality inspection methods are generally used at present. One is to use probability statistics to conduct manual sampling quality inspection. Missing data or low precision will directly affect the accuracy of quality inspection results

The other is to use rules or matching algorithms to machine-recognize and score voice data. This method can realize the quality inspection of all voice data one by one, but the rules used in this method are relatively fixed and rigid. Approximate semantic expressions are difficult to match effectively

It can be seen that in the field of voice quality inspection, there is still a lack of a quality inspection scheme that can not only ensure the full coverage of voice data, but also accurately detect whether the semantic information in the voice data is qualified.

Images