Data labeling method and device based on self-learning algorithm

Abstract

The invention relates to the field of voice signal processing, in particular to a data labeling method and device based on self-learning algorithm. The method comprises a speech recognition step, a text comparison step, a natural language processing algorithm evaluation step, a natural language processing algorithm prediction step, a data labeling step, a quality inspection step and a self-learning step. The text comparison step is used for comparing a plurality of recognition texts, labeling difference parts of texts and performing sentence breaking processing. The data labeling step is usedfor performing data labeling on an optimal pre-labeled text for a plurality of times by referring to an original recognition text and a prediction text of the difference parts, so as to form a plurality of groups of data labeling texts. The self-learning step is used for inputting the optimal labeled text and a corresponding audio signal into a speech recognition engine, wherein the speech recognition engine is iteratively trained based on the self-learning algorithm. According to the labeling method and device, the data labeling time is greatly saved, the data labeling quality and the data labeling efficiency are effectively improved, the training support is provided for various artificial intelligence products, and the production effect of intelligent products is improved.

Description

technical field [0001] The invention relates to the field of speech signal processing, and more specifically, to a data labeling method and device based on a self-learning algorithm. Background technique [0002] With the development of artificial intelligence (AI, Artificial Intelligence) technology, intelligent products such as voice robots, agent assistants, and voice quality inspection based on ASR technology have been widely promoted. ASR, the full name in English is AutomatedSpeech Recognition, that is, automatic speech recognition technology, which is a technology that converts human speech into text. The accuracy of ASR will directly affect the effect of smart products. [0003] A self-learning algorithm must have a large number of training sets and test sets. The data in the test set and training set must be labeled data that supports the algorithm. The process of converting the collected raw data into data available to the algorithm is called data labeling. That ...

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to provide a data labeling method and device based on a self-learning algorithm to solve the problems of low intelligence, long manpower hours, low efficiency and poor quality in current data labeling

Images