Customer service voice quality inspection method based on text classification

Abstract

The invention discloses a customer service voice quality inspection method based on text classification, and the method is characterized in that the method comprises the steps: 1, designing a databasetable structure, and developing ETL; step 2, identifying the telephone voice by using voice identification software; step 3, training an identity recognition model, and then recognizing the identityof the text; 4, performing sensitive word matching on the texts distinguished from the customer service and the customer identities, and writing a matching result into a database; and 5, carrying outdata storage and regular scheduling. Quality inspection personnel can be effectively helped to improve the efficiency, and manpower and time cost are saved; on the premise of a high-accuracy voice recognition technology, customer service is made, customer identity recognition is carried out, quality inspection personnel carry out manual examination on customer service speaking content with sensitive words recognized by the system, and spot check is carried out on the sensitive content which is not recognized; and meanwhile, the dialogue which has been manually audited is reserved as corpus, sothat a basis is provided for later attempt to carry out sensitive detection on the text based on the semantic model.

Description

Technical field [0001] The invention relates to the field of speech quality inspection, and in particular to a customer service speech quality inspection method based on text classification. Background technique [0002] Generally speaking, the outbound voice of customer service should be retained forever. Many industries have dual recording requirements for permanent reference. Take the insurance industry as an example, when there is a dispute or complaint, the supervisory committee will require checking whether there is any sensitivity in the telemarketing call. Therefore, it is important to keep the telephone return visit information of insurance customers and assist in the settlement of claims. For a long time, the work method of customer service voice quality inspection is mainly based on manual random inspections, relying on customer service to listen to voices, watch chat records with customers, and combine their own professional judgments to conduct compliance inspections...

AI Technical Summary

Problems solved by technology

For a long time, the working method of customer service voice quality inspection has been mainly based on manual spot checks, relying on customer service to listen to voice, read chat records with customers, and combine their own professional judgments to conduct compliance quality inspections. The disadvantages of traditional voice quality inspection methods (the ratio of manual sampling is very low, which means that a large number of fixed-line voices are ignored, including hidden valuable or risky voices that have not been discovered) are clearly exposed

Images