Text scoring method, text scoring model, text scoring device and storage medium

Abstract

The embodiment of the invention provides a text scoring method, a text scoring model, text scoring equipment and a storage medium, and the text scoring method comprises the following steps: dividing an input text according to a preset division granularity to obtain a plurality of text segments; based on each text fragment, carrying out data coding to obtain corresponding fragment coding data; carrying out data coding based on the input text to obtain sequence coding data; wherein the sequence coding data comprises text coding sub-data and / or a plurality of language element coding sub-data; and based on the fragment coding data corresponding to the plurality of text fragments and the sequence coding data, performing score prediction to obtain a text prediction score. By adopting the scheme, the score prediction accuracy can be improved.

Description

technical field [0001] The embodiments of this specification relate to the technical field of natural language processing, and in particular to a text scoring method, a text scoring model, a text scoring device, and a storage medium. Background technique [0002] Traditional composition grading is done manually, but manual grading requires a lot of manpower, and manual grading is highly subjective. With the development of computer science and technology, the ability and level of automated information processing has also been significantly improved, and Automated Essay Scoring (AES) technology has emerged as the times require. [0003] AES technology can be mainly divided into traditional AES methods and AES methods based on deep learning. Among them, the traditional AES method manually extracts the scoring-related data in the text, and uses simple neural network models such as regression models, classification models, or ranking models to predict scores. Because the tradit...

AI Technical Summary

Problems solved by technology

[0005] However, if there are too many punctuation marks, the divided text will become fragmented, and for long sentences, due to the length of the sentence is too long, some feature information will be lost during the processing process. Therefore, in the face of complex and changeable text, the accuracy of score prediction is still low

Therefore, the existing AES method has the problem of poor score prediction

Images