Data processing method and device

Abstract

The invention provides a data processing method and device. The data processing method comprises the following steps of obtaining training data, wherein the training data includes voice data and corresponding labeling results; generating a first acoustical model according to the training data; using the first acoustical model to perform voice recognition on the training data to obtain a recognition result; determining the affiliated type of the training data according to the labeling result and the recognition result. The method has the advantages that the training data can be automatically classified, and the problems of a manual random inspection mode can be avoided.

Description

technical field [0001] The present application relates to the technical field of speech recognition, in particular to a data processing method and device. Background technique [0002] The performance of the acoustic model will have a great impact on the performance of speech recognition, and the scale and quality of the training data will affect the performance of the acoustic model. Especially when there is a large-scale training data, how to control the quality of the training data becomes an urgent problem to be solved. [0003] The training data used for acoustic model training usually includes speech data and corresponding labeled data, and the labeled data is usually manually labeled. However, there will be a certain error rate in manual labeling, and it is not required to be all correct even during acceptance, for example, an accuracy rate of more than 95% is required. Therefore, the training data will contain certain mislabeled data. In addition, there will be da...

AI Technical Summary

Problems solved by technology

However, the manual sampling method has problems such as time-consuming and labor-intensive

Images