Chinese error correction model training method and Chinese error correction method and device

Abstract

The embodiment of the invention provides a Chinese error correction model training method and a Chinese error correction method and device, and the method comprises the steps: pre-training an initialmodel based on a first training data set, and obtaining a first pre-training model; finely adjusting the first pre-training model based on the first training data set and a second training data set toobtain a Chinese error correction model, wherein the first training data set comprises a plurality of sample quadruples, and each sample quadruple is composed of four elements including a Chinese corpus, a pinyin sequence, a stroke sequence and a picture sequence, the second training data set comprises a plurality of similar sample quadruples, and the similar sample quadruples are obtained by replacing characters corresponding to similar characters in any Chinese corpus based on the similar characters in the Chinese corpus. According to the training method of the Chinese error correction model and the Chinese error correction method and device provided by the embodiment of the invention, the generation efficiency and richness of the training data are improved, and the error correction effect of the trained Chinese error correction model is relatively good.

Description

technical field [0001] The present invention relates to the technical field of natural language processing, in particular to a training method for a Chinese error correction model, a Chinese error correction method and a device. Background technique [0002] Chinese error correction is a very important topic in natural language processing. The error correction model automatically finds and corrects potential grammatical errors in the text, and keeps the corrected semantics unchanged. The error correction model can not only improve the quality of the text, It also makes it easier for downstream modules in your program to process text. [0003] However, the construction of error correction models faces some challenges. The current construction of error correction models often refers to the training methods of machine translation. The input of the machine translation model is the original sentence in one language, and the output is the translated sentence in another language a...

AI Technical Summary

Problems solved by technology

[0005] Embodiments of the present invention provide a training method for a Chinese error correction model, a Chinese error correction method and a device to solve the defects in the prior art of high cost of artificial marking and poor adaptability of the error correction model to Chinese

Images