Automatic essay scoring (AES) refers to the use of computers to automatically score essays without human intervention. Whether the discourse structure is reasonable is an important consideration dimension in narrative scoring. At present, the research on text structure at home and abroad is still in its infancy, and there is a lack of open corpus for training. In addition, most of the text structure evaluation models constructed in the past are trained by selecting features based on manual experience. To solve the above problems, this paper develops text composition standards and labeling norms, supplements the previously constructed Chinese narrative text structure assessment corpus, and puts forward a narrative text structure assessment model with the aid of text composition. The experimental results show that the model constructed in this paper can effectively identify text components, and the accuracy of the narrative text structure assessment model with the help of text components is better than that of the baseline model, with an accuracy of 76.8%.