Automatic speech quality testing method and system for calling system

Abstract

The invention provides an automatic speech quality testing method for a calling system. The method comprises the steps that a speech data acquisition module obtains speech data from the calling system, then the acquired speech data is sent to a text-speech translation module, and the text-speech translation module converts the speech data into text data and sends the text data obtained after translation to a text data storage module for saving; and an intelligent quality testing condition generation module obtains intelligent quality testing conditions according to quality testing data of an intelligent quality testing module and a quality testing learning algorithm. Through automatic speech quality testing method for the calling system, speech quality testing can be realized systematically and intelligently, full-amount coverage quality testing can be performed, quality testing efficiency is high, the workload of quality testing staff is lowered, the quality testing staff concentrate on a small amount of data not passing quality testing due to a low quality testing score and quality testing fuzzy data, and meanwhile multiple quality testing conditions are quickly fused to perform quality testing evaluation on a speech at the same time.

Description

Technical field [0001] The present invention relates to the technical field of calling systems, in particular to a voice automatic quality inspection method of a calling system. Background technique [0002] The management system formed by the computer network has become information production and management. The customer service personnel handle tens of thousands of calls every day, and the quality inspection personnel can only conduct random inspections of these call recordings through manual audiometry. The rate of audiometry sampling is low and quality inspection The workload is large, the efficiency is low and the coverage is low, and it is difficult to effectively evaluate the overall service quality. Sometimes it may be a very vague recording. It may not always have a major problem. It may be that some expressions are not good, and there are expressions that should not appear in the expression. These are constantly being tested manually. Repeatedly listen to a recording. ...

AI Technical Summary

Problems solved by technology

[0002] The management system formed through the computer network has become information production and management. Customer service personnel handle tens of thousands of calls every day, and quality inspection personnel can only conduct spot checks on these call recordings through manual audiometry. Heavy workload, low efficiency and low coverage, it is difficult to effectively evaluate the overall service quality

Sometimes it may be a very vague recording, not necessarily a major problem, it may be that some expression methods are not good, and there are words that should not appear in the expression. These are continuously tested manually, and it may take Repeatedly listening to a recording, the existing manual voice detection method is inefficient and has a low coverage rate, so manual detection is greatly affected by human factors, and some key words may be missed in a certain period of time , it is impossible to accurately detect bad recordings, or use random inspection to calculate a certain probability to detect, which will also cause some problems, manual random inspection, there will always be some cases that cannot be detected, so some companies change the detection to fuzzy Inspection, although such inspection is efficient, it cannot strictly complete the quality inspection;

Images