基于prompt的文本可读性评估  

Prompt-based text readability assessment

作　　者：冯宇涛 强继朋 李云[1] 袁运浩[1] 朱毅 FENG Yutao;QIANG Jipeng;LI Yun;YUAN Yunhao;ZHU Yi(School of Information Engineering,Yangzhou University,Yangzhou 225127,China)

机构地区：[1]扬州大学信息工程学院,江苏扬州225127

出　　处：《扬州大学学报（自然科学版）》2023年第3期56-60,共5页Journal of Yangzhou University：Natural Science Edition

基　　金：国家自然科学基金资助项目(62076217,61906060);扬州大学“青蓝工程”资助项目。

摘　　要：为了提高预训练模型在文本可读性分类中的表现,提出基于预训练模型BERT(bidirectional encoder representations from transformers)和GPT2(generative pre-training 2)的prompt文本可读性分类方法,通过手工模板和映射器将文本可读性分类转换为与模型预训练任务一致的文本填空形式,以此提高预训练模型的分类性能.实验表明,prompt方法在实验数据集上与传统机器学习和有监督的神经网络相比具有更好的分类性能.In order to improve the performance of the pre-trained models in text readability classification,this paper proposes a prompt text readability classification method based on the pre-trained models BERT(bidirectional encoder representations from transformers)and GPT2(generative pretraining 2).This method converts text readability classification into text filling form consistent with model pre-training task by manual template and verbalizer,so as to improve the classification performance of pre-training model.Experiments show that the prompt methods achieve better classification performance on the experiment datasets than traditional machine learning and supervised neural networks.

关 键 词：文本可读性分类 预训练模型 手工模板 

分 类 号：TP391.12[自动化与计算机技术—计算机应用技术]