Chinese Text Correction System Based on Ernie

Nowadays, Internet technology is developing at a high speed, which makes the vast number of Internet users constantly producing and receiving huge amounts of information data. Because of the low threshold of Internet use, it leads to the uneven quality of information on the Internet, and the quality of most of the information is decreasing. The quality of information can be improved by proofreading the text for errors. Therefore, it is important to detect and correct typos in text information. With the more widespread application of deep learning in the field of natural language processing, a deep learning and rule-based approach to Chinese text detection and error correction has achieved better results in processing text compared to traditional methods. The main work is as follows: the existing deep learning algorithms and models on text error correction are compiled, briefly introduced and compared, in which the effect of Ernie and Bert-based Chinese error correction processing is mainly compared, laying the foundation for selecting the ernie end-to-end Chinese error correction model. The error correction process of the Ernie model is also introduced, and the prediction results are evaluated and trained on the selected dataset. After experimental validation, the Ernie-based end-to-end model achieves better results with a greater improvement in accuracy. Moreover, the front-end interface design was carried out, and the Chinese text error correction system was successfully built, which can perform functions such as uploading and downloading of files, and support the saving of results.