Method and system for optimizing speech recognition acoustic model, equipment and storage media

Abstract

Embodiments of the present invention disclose a method and a system for optimizing a speech recognition acoustic model, equipment and a storage media. The method includes: obtaining a labeled text ofa sample speech, and acquiring an identification text obtained by the sample speech based on the current acoustic model; comparing the labeled text with the identification text, and determining errorlabeling information of the labeled text relative to the identification text when a comparison result is not matched; updating the labeled text of the sample speech according to a text update decisioncondition corresponding to the error labeling information; and retraining and optimizing the current acoustic model based on sample speeches of the set amount and current corresponding labeled texts.By using this method, the labeling quality of the labeled text corresponding to the sample speech can be effectively improved, thereby achieving the purpose of optimizing the acoustic model.

Description

technical field [0001] The invention relates to the field of computer technology, in particular to a method, system, device and storage medium for optimizing an acoustic model for speech recognition. Background technique [0002] With the continuous expansion of the application range of speech recognition, speech recognition technology has become a new high-tech industry, and has attracted more and more technical personnel's attention. At present, one of the important components in the speech recognition system is the acoustic model. The quality of the acoustic model largely determines the quality of the speech recognition results. Therefore, it is necessary to continuously optimize the speech recognition acoustic model. [0003] Generally, the training of the acoustic model requires the support of a large amount of sample data, and the sample data often includes voice data and marked text corresponding to the voice data (text content included in the voice data). Annotated ...

AI Technical Summary

Problems solved by technology

Annotated text is usually achieved based on a large number of manual annotations or obtained through identification by a third-party recognition system. However, there are often certain errors in the annotated text obtained through the above methods, which affect the quality of annotations.

[0004] For the speech recognition acoustic model, improving the annotation quality of the annotation text is equivalent to one of the means to optimize the acoustic model, but there is no technical solution to realize the optimization of the acoustic model by improving the quality of the annotation text.

Images