Text data generation method, computer equipment and storage medium

Abstract

The embodiment of the invention provides a text data generation method, computer equipment and a storage medium. The text data generation method comprises the steps of obtaining a third text without grammar defects; the third text is input into a generation model, a first generation text corresponding to the third text and a target probability used for indicating that the first generation text has grammatical defects are generated, and the generation model is obtained through training based on a first text without grammatical defects and a second text with grammatical defects; and determining the first generated text as a target text according to the target probability corresponding to the first generated text. The generation model learns the grammar error rule corresponding to the second text, the trained generation model can play the excellent characteristic of diverse generation results, and a large number of texts containing grammar defects can be automatically constructed.

Description

technical field [0001] The present application relates to the technical field of natural language processing, and in particular to a method for generating text data, computer equipment and storage media. Background technique [0002] The goal of the grammatical error detection task is to detect possible grammatical errors in the text. These grammatical errors can be divided into "typos", "incomplete components", "redundant components", "improper collocations", "improper words", " Improper word order" and other categories. The task of grammatical error detection requires the use of text marked with grammatical errors for model training. However, the current Chinese corpus marked with grammatical errors is small in scale, difficult to obtain, and expensive to label. [0003] Due to the lack of enough labeled data, the grammatical error detection model is prone to overfitting on small-scale training sets, which seriously restricts the improvement of the grammatical error dete...

AI Technical Summary

Problems solved by technology

[0003] Due to the lack of enough labeled data, the grammatical error detection model is prone to overfitting on small-scale training sets, which seriously restricts the improvement of the grammatical error detection model

On the other hand, the small scale of labeled data also means that there are fewer grammatical errors in the data, and there may even be extreme cases where most of the errors in the data are misuses of "de, get, and place", which leads to grammatical errors in the data. Can be trained on few error types, it is difficult to learn enough general knowledge

Images