The mistake detection and correction system is crucial in the pronunciation of spoken English, yet it has a problem with erroneous performance placement. The usual genetic algorithm fails miserably when it comes to correcting the positioning of incorrect corrections in spoken English pronunciation. This study concludes with the provision of an assessment of the research on an English speech mistake detection and correction system based on artificial neural networks. First, in order to reduce interference factors in the error detection and correction system, the indicators are separated according to the demands of the system, and the influencing components are discovered using gradient descent theory. We then use gradient descent theory to the design of an artificial neural network (ANN) error detection and correction system, and we analyze the results in detail. In terms of accuracy of error detection and correction systems and time of influencing variables, the MATLAB simulation results show that artificial neural networks beat the standard genetic algorithm under specific evaluation conditions.