摘要
近年来中文预训练模型层出不穷,在多数中文分类和生成任务上达到了最优的性能,且随着预训练模型的不断增大,如何充分利用预训练的所有权重也成为了一个热门的课题。文章结合中文多目标情感分类任务,提出了基于生成式模型的情感分类模型,在预训练模型的基础上,充分利用预训练权重,并基于一定的范式引导模型利用生成模型预测多目标的情感类别。实验证明,使用预训练的生成模型和人工制定的范式进行多目标情感分类时准确率达到了最优水平。
In recent years,Chinese pre-training models emerge in endlessly,achieving the optimal performance in most Chinese classification and generation tasks.With the continuous increase of pre training models,how to make full use of all weights of pre-training has become a hot topic.In this paper,combined with the Chinese multi-objective emotion classification task,we propose an emotion classification model based on the generative model.On the basis of the pre-training model,we make full use of the pre-training weight,and use the generative model to predict the multi-objective emotion category based on a certain paradigm guidance model.The experiment proves that the accuracy of multi-objective emotion classification using pre trained generation model and manually formulated paradigm reaches the optimal level.
作者
王艳
任晓航
张秋旋
邢晓岑
王铭康
WANG Yan;REN Xiao-hang;ZHANG Qiu-xuan;XING Xiao-cen;WANG Ming-kang(Space Engineering University,Beijing 102299,China)
出处
《电脑与信息技术》
2023年第5期1-4,26,共5页
Computer and Information Technology
关键词
多目标
情感分类
生成式模型
深度学习
multi objective
emotional classification
generative model
Deep learning
