Text data generation method for training English grammar error correction model

The invention relates to the technical field of data generation, in particular to a text data generation method for training an English grammar error correction model. The text data generation methodcomprises the following steps: (1) introducing sentence error quantity; (2) determining an error type...

Full description

Saved in:
Bibliographic Details
Main Authors QIN LONG, CHEN JIN, XU SHUYAO
Format Patent
LanguageChinese
English
Published 01.11.2019
Subjects
CALCULATING
COMPUTING
COUNTING
HANDLING RECORD CARRIERS
PHYSICS
PRESENTATION OF DATA
RECOGNITION OF DATA
RECORD CARRIERS
Online AccessGet full text

Cover

More Information
Summary:The invention relates to the technical field of data generation, in particular to a text data generation method for training an English grammar error correction model. The text data generation methodcomprises the following steps: (1) introducing sentence error quantity; (2) determining an error type; (3) carrying out corresponding Word Tree replacement according to the error type; and (4) utilizing WMT11 single-language data and One-Billion-Word single-language data to generate grammar correction model pre-training data. According to the text data generation method, the effect of the grammarcorrection model is effectively improved. 本发明涉及数据生成技术领域，尤其是一种用于训练英语语法改错模型的文本数据生成方法，其步骤为：(1)句子错误数量的引入；(2)确定错误类型；(3)根据错误类型进行相应的Word Tree替换；(4)利用WMT11单语言数据与One-Billion-Word单语言数据生成作为语法改错模型预训练数据，本发明有效提高语法改错模型的效果。
Bibliography:Application Number: CN201910719334