Voice data automatic annotation quality evaluation method

Abstract

The invention provides a voice data automatic annotation quality evaluation method, which comprises the steps of pre-constructing a quality rule base for automatically annotating voice data based on aquality key index; reading to-be-detected automatic annotation voice data, and performing quality detection on the to-be-detected automatic annotation voice data according to the quality key indexesto complete quality measurement; updating an automatic annotation voice data set according to the result of the quality measurement; and converting the updated automatic annotation voice data set intoa new rule, and importing the new rule into the quality rule base. According to the method, the defects existing when a traditional data annotation quality evaluation method is used for automaticallyannotating data by a machine are overcome; and the method has a very positive support effect on promoting the intelligent development process of the small language voice.

Description

technical field [0001] The invention relates to the technical field of language information processing, in particular to a quality assessment method for automatic labeling of voice data. Background technique [0002] In recent years, automatic data labeling has gradually become a key basic technology in the field of artificial intelligence. It is hoped that automatic labeling of data by machines will replace manpower, and great progress has been made in automatic data labeling in fields such as images. The extreme scarcity of speech annotation data has become a key factor restricting the speech recognition performance of minority languages ​​in my country. Due to the influence of original data quality, human errors, and model limitations, data labeling errors are unavoidable. Therefore, it is very important to introduce effective quality assessment methods. However, data labeling standards are not uniform, and labeling quality is uneven. To a large extent, it hinders the ap...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to solve the defects in the above-mentioned prior art, and provide a quality assessment method for automatic voice data labeling, which solves the following problems: first, perform quality assessment on the voice automatic labeling data completed by the machine, and find that the labeling data exists "Mislabeling and missing labeling" and other quality problems, so as to improve the quality of automatic data labeling

Images