Evaluation of Chinese Narrative Text Structure by Integrating Text Component Identification
The rationality of the text structure is an important dimension in automatic essays scoring.The research on automatic evalua-tion of the text structure in narrative essay is still in its infancy.In this paper,by working with experts to formulate text structure evalu-ation criteria and text component labeling specifications,a large-scale Chinese narrative text structure evaluation corpus is constructed.Based on this corpus,this paper proposes an evaluation model of narrative text structure that integrates text component identification.The model uses deep learning algorithm and attention mechanism to learn text features from the three levels of words,sentences and paragraphs,so as to extract important information about the text structure,and finally score the structural rationality by integrating the identified text components.The experimental results on the constructed narrative text structure corpus show that the accuracy of the proposed model is better than the existing work and baseline models,reaching 79.6％.
text structurecorpusautomatic essay scoringhierarchical networkBERT
NETL
NSTL
万方数据
