ANALYSIS OF PRIMARY SCHOOL STUDENTS' ASSESSMENT TEXT BASED ON INTELLIGENT TECHNOLOGY: AUTOMATIC CLASSIFICATION OF COGNITION, EMOTION AND REFLECTION

Abstract

Digital transformation of the peer assessment process based on intelligent technology is the key to building a new student assessment ecosystem. As the key process data of peer assessment, the coding of peer assessment text is still limited by the level and time of professionals. In order to improve the speed of text encoding, realize automatic classification and instant feedback, artificial intelligence technology is used to automatically encode cognition, emotion and reflection in 10,771 primary school students' peer assessment text using the ALBERT deep learning model. The study found that the model implements semantic-based text quality evaluation and can perform multi-dimensional hierarchical evaluation of the FCR model in the peer assessment text. Compared with other classic models, it has better classification effect. The accuracy of different dimensions ranges from 78.78% to 97.23%. This study constructs a comprehensive, fast and accurate peer assessment method, which lays an important foundation for building an assessment model based on the teaching process and large-scale intelligent student assessment.