Design of Chinese pronunciation correction education game based on improved twin network pronunciation algorithm

Abstract

Against the backdrop of China’s rising international standing, enthusiasm for Chinese language learning among people across the globe has been continuously surging. To lower the barriers to Chinese learning, this paper proposes a Chinese education game with pronunciation correction functionality, which is built on an improved Siamese network and a dedicated pronunciation detection algorithm. To verify the performance of the proposed algorithm, this study conducts comparative experiments with the conventional pronunciation proficiency evaluation algorithm and the weight-allocation-based Siamese convolutional network. The experimental results show that the accuracy, recall, and F1 means of speech detection based on twin convolutional neural networks with limited weight sharing are 89.5 %, 80.5 %, and 0.81, respectively, which are higher than the other two algorithms. This indicates the excellent performance of this algorithm. Meanwhile, in the verification of learning effect, the average score of those who learned through Chinese education games that can correct pronunciation was 74 points, higher than that of those who learned through conventional methods. The above results indicate that improving the twin network pronunciation detection algorithm for detecting learners’ pronunciation can effectively improve their learning effectiveness.