Automatic Assessment System for English Pronunciation Machine Quality Based on GloVe-CNN Algorithm

In recent years, people have paid more and more attention to machine quality assessment of pronunciation, and some progress has been made. For a long time in the past, people only paid attention to the symbols and phonetic information in mechanical language, but did not pay attention to whether the relationship between multiple corpus materials was accurate and reliable. Moreover, with the rapid development of computer technology, network communication technology, and the continuous maturity and improvement of digital coding research methods and theories, the English pronunciation machine evaluation system is also constantly being updated, and the quality evaluation level of pronunciation machines is getting higher and higher. This article aims to explore this method's assessment of users' English proficiency from three aspects: speech quality, grammatical structure and semantic features. First, this article introduces the problem of automatic assessment of speech quality and its related influencing factors; secondly, elaborates on the common measurement standards for judging whether the speaker is “Chinese-English bilingual“; finally, introduces the data mining result model based on machine learning algorithms, and explains the function used to construct sentence fuzzy sets. The overall accuracy of the system in this article is over 94%. Therefore, the system has an effective mechanism to detect and correct learners' errors in pronunciation of phonetic symbols. The system uses convolutional neural network to learn the complex relationship between pronunciation quality and features. CNN can effectively learn different levels of features and extract meaningful information for pronunciation quality assessment.