Online spoken language pronunciation evaluation method and device and storage medium

Abstract

The invention provides an online spoken language pronunciation evaluation method and device and a storage medium. The method comprises the steps of obtaining a target text, a pronunciation dictionary and a read audio made by a user according to the target text; by using a pre-trained speech recognition model, on one hand, decoding target speech data to generate a phoneme decoding result, and on the other hand, carrying out forced alignment processing based on the target text on the target speech data to obtain a word and phoneme alignment result; and according to the phoneme decoding result and the word and phoneme alignment result, obtaining pronunciation accuracy scores of the read audio under different voice granularities, the voice granularities comprising a phoneme level, a word level and a sentence level. According to the method, the spoken language of a language learner is scored and the score is fed back in real time by using a voice evaluation technology, the language learner is helped to complete high-quality learning in a closed-loop learning environment of continuous trial and error, and the spoken language ability of the user is improved.

Description

technical field [0001] The present invention relates to the technical field of speech recognition and evaluation, in particular, to an online oral pronunciation evaluation method, device and storage medium. Background technique [0002] Voice evaluation is a software system that automatically evaluates pronunciation level, pronunciation errors, defect location and problem analysis through intelligent voice technology. [0003] The existing oral speech quality evaluation is generally achieved by comparing the characteristic parameters of the test speech with the characteristic parameters of the standard speech, which cannot be evaluated at the granularity of small pronunciation units. The evaluation results will also be affected by the speaker's personal pronunciation characteristics. For example, if the speaker and the standard pronunciation reader are of the same gender and age, the score will be higher. Another method is to achieve pronunciation evaluation results by extr...

AI Technical Summary

Problems solved by technology

Another method is to achieve pronunciation evaluation results by extracting phonetic features for pattern comparison, but the obtained evaluation results cannot exclude the influence of the speaker's personal pronunciation characteristics

In addition, this method uses speech recognition technology to obtain speech recognition text results, and then compares the recognition text with the target text, and cannot perform quality evaluation at the granularity of small pronunciation units.

Images