Design and Implementation of Language Automatic Error Correction System Based on Convolutional Neural Network and Human-Computer Interaction

In view of the problem that the current grammatical error correction (GEC) system has poor grammatical error handling in complex sentence structures and is difficult to provide personalized grammar correction user experience, this paper combines convolutional neural network (CNN) and Human-computer interaction (HCI) design to study the design and implementation of English grammar automatic correction system, aiming to improve the accuracy and user experience of GEC in complex sentence structures. First, a pre-trained word embedding model is used to map each word into a vector representation of a fixed dimension and construct a sentence matrix representation. Then, based on CNN, local features in the text are extracted through convolution operations to identify grammatical errors and classify them. Finally, a simple and intuitive HCI interface is designed to ensure that the interface is intuitive, easy to use and can promote efficient user participation. The experimental results show that compared with the Transformer-based and rule-based GEC systems, the average F0.5 scores of this system in the CoNLL-2014 (Conference on Natural Language Learning-2014) and Lang-8 public dataset tests are 3.92% and 6.54% higher and 3.42% and 7.04% higher, respectively. The conclusion shows that this system helps to improve the performance of GEC and provide language learners with intelligent and easy-to-use learning tools.