Automatic generation method for public opinion sentiment text based on large language generation model

CUI Su; HAN Yiliang; CHEN Yali; ZHU Shuaishuai

doi:10.11959/j.issn.2096-109x.2025074

您当前的位置：

首页 >

文章列表页 >

Automatic generation method for public opinion sentiment text based on large language generation model

Papers | 更新时间：2026-01-28

- Automatic generation method for public opinion sentiment text based on large language generation model
- Chinese Journal of Network and Information Security Vol. 11, Issue 6, Pages: 157-164(2025)
- 作者机构：
  
  1.中国人民武装警察部队工程大学，陕西西安 710086
  2.中国人民武装警察部队警官学院，四川成都 610000
- 作者简介：
- 基金信息：
  
  The Natural Science Basis Research Plan in Shaanxi Province of China(2024JC-YBMS-546)
- DOI：10.11959/j.issn.2096-109x.2025074
  CLC： TP391
- Received：01 February 2025，
  
  Revised：2025-07-10，
  
  Published：25 December 2025
- 稿件说明：
移动端阅览
崔骕, 韩益亮, 陈雅丽, 等. 基于大语言生成模型的舆情情感文本自动生成方法[J]. 网络与信息安全学报, 2025, 11(6): 157-164.

CUI Su, HAN Yiliang, CHEN Yali, et al. Automatic generation method for public opinion sentiment text based on large language generation model[J]. Chinese Journal of Network and Information Security, 2025, 11(6): 157-164.
崔骕, 韩益亮, 陈雅丽, 等. 基于大语言生成模型的舆情情感文本自动生成方法[J]. 网络与信息安全学报, 2025, 11(6): 157-164. DOI： 10.11959/j.issn.2096-109x.2025074.

CUI Su, HAN Yiliang, CHEN Yali, et al. Automatic generation method for public opinion sentiment text based on large language generation model[J]. Chinese Journal of Network and Information Security, 2025, 11(6): 157-164. DOI： 10.11959/j.issn.2096-109x.2025074.

摘要

社交网络中的负面舆情事件在传播过程中会产生巨大的消极影响。为快速自动生成正面舆情文本，化解消极情绪、缓解舆论压力，提出了一种基于大语言生成模型（ERNIE-GEN）的舆情情感文本生成方法。首先，基于RoBERTa模型构建情感分类器，对舆情文本进行情感极性识别与分类，得到正面、中性和负面情绪的文本；然后，利用TextRank算法提取不同情感文本的关键主题，刻画当前主题对应的舆情事件；接下来，将正面主题和情感文本输入ERNIE-GEN模型进行训练，得到正面积极舆情文本生成模型；最后，将负面舆情事件的主题词输入该模型，即可生成对应的正面积极舆情文本。以“某医院暴力伤医案”期间“医生”主题舆情事件为例，通过模型生成与负面主题相对的正面积极情感文本，用于缓解和稀释热点舆情，转移公众注意力。实验结果显示，所提方法在网络舆情引导方面具有良好的效果。

Abstract

Social media negative public opinion events can have a tremendous adverse impact during their dissemination process. To intelligently and rapidly generate positive public opinion texts for resolving negative emotions and alleviating public opinion pressure

a new method of generating sentiment texts based on the large language generation model (ERNIE-GEN) was proposed. Firstly

an emotion classifier was constructed based on RoBERTa model to identify and classify the emotion polarity of public opinion texts

and obtain positive

neutral and negative emotion texts. Then

the TextRank algorithm is used to extract the key themes of different emotional texts and depict the public opinion events corresponding to the current theme. Next

the positive theme and emotion text were put into the ERNE-GEN model for training

and the positive public opinion text generation model was obtained. Finally

the topic words of negative public opinion events were input into the model to generate positive public opinion event text. Taking the "doctor" theme public opinion event during the "violent assault on doctors incident at a certain hospital" as an example

positive public opinion event texts were generated according to the negative public opinion event theme

which were used to alleviate and dilute hot public opinion and divert public attention. The experimental results show that the proposed method has a very good effect in guiding online public opinion.

关键词

Keywords

references

CHEN Z T , LAN H L . Dynamics of public opinion: diverse media and audiences’ choices [J ] . Journal of Artificial Societies and Social Simulation , 2021 , 24 ( 2 ): 8 .

ZHANG L J , WANG T , JIN Z L , et al . The research on social networks public opinion propagation influence models and its controllability [J ] . China Communications , 2018 , 15 ( 7 ): 98 - 110 .

余晓青 . 意识形态网络舆情应对机制研究 [J ] . 南京邮电大学学报(社会科学版) , 2021 , 23 ( 5 ): 37 - 47 .

YU X Q . The problems and countermeasures of coping mechanism on ideological internet public opinion [J ] . Journal of Nanjing University of Posts and Telecommunications (Social Science) , 2021 , 23 ( 5 ): 37 - 47 .

刘淑婷 , 马颖华 , 陈秀真 . 面向舆情治理的信息管理机制演化博弈模型 [J ] . 网络与信息安全学报 , 2023 , 9 ( 6 ): 102 - 115 .

LIU S T , MA Y H , CHEN X Z . Evolutionary game model of information management mechanism for public opinion governance [J ] . Chinese Journal of Network and Information Security , 2023 , 9 ( 6 ): 102 - 115 .

胡益 , 刘嘉勇 , 代金鞘 , 等 . 结合主题模型与自监督学习的可控文本生成技术研究 [J ] . 四川大学学报(自然科学版) , 2023 , 60 ( 5 ): 130 - 138 .

HU Y , LIU J Y , DAI J Q , et al . Controllable text generation technology based on topic model and self supervised learning [J ] . Journal of Sichuan University (Natural Science Edition) , 2023 , 60 ( 5 ): 130 - 138 .

阮叶丽 . 基于GAN考虑外部知识与情感属性的文本生成研究 [D ] . 武汉 : 中南财经政法大学 , 2021 .

RUAN Y L . The research of text generation with considering external knowledge and emotional attributes based on GAN [D ] . Wuhan : Zhongnan University of Economics and Law , 2021 .

WANG Y Q , WU R , ZENG J , et al . Research on the public opinion guidance mechanism of major public health incidents [J ] . Frontiers in Psychology , 2022 , 13 : 872464 .

WEI J , JIA Y G , TIE W Y , et al . Opinion evolution with information quality of public person and mass acceptance threshold [J ] . Big Data , 2024 , 12 ( 2 ): 100 - 109 .

YOU G , GAN S Q , GUO H , et al . Public opinion spread and guidance strategy under COVID-19: a SIS model analysis [J ] . Axioms , 2022 , 11 ( 6 ): 296 .

刘欣然 , 徐雅斌 , 李继先 . 网评贴文自动生成方法研究 [J ] . 数据分析与知识发现 , 2023 ( 4 ): 101 - 113 .

LIU X R , XU Y B , LI J X . Method for automatically generating online comments [J ] . Data Analysis and Knowledge Discovery , 2023 ( 4 ): 101 - 113 .

王沛 , 杨频 , 程芃森 , 等 . 基于外部知识筛选的主题文本生成技术研究 [J ] . 四川大学学报(自然科学版) , 2024 , 61 ( 1 ): 75 - 83 .

WANG P , YANG P , CHENG P S , et al . Research on topic text generation technology based on external knowledge filtering [J ] . Journal of Sichuan University (Natural Science Edition) , 2024 , 61 ( 1 ): 75 - 83 .

LYU B Y , DU Y J , JIANG J J , et al . Positive public opinion guidance model based on dual learning in social network [J ] . Intelligent Data Analysis , 2023 , 27 ( 3 ): 833 - 853 .

SONG J B , ZHU X Y . Research on public opinion guidance of converging media based on AHP and transmission dynamics [J ] . Mathematical Biosciences and Engineering , 2021 , 18 ( 5 ): 6857 - 6886 .

LIU Y H , OTT M , GOYAL N , et al . RoBERTa: a robustly optimized BERT pretraining approach [J ] . arXiv preprint arXiv:1907. 11692 , 2019 .

XIAO D L , ZHANG H , LI Y K , et al . ERNIE-GEN: an enhanced multi-flow pre-training and fine-tuning framework for natural language generation [C ] // Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence . International Joint Conferences on Artificial Intelligence Organization , 2020 : 3997 - 4003 .

World Health Organization . Coronavirus disease (COVID-19) - events as they happen [EB ] . 2020 .

CLARK K , LUONG M T , LE Q V , et al . ELECTRA: pre-training text encoders as discriminators rather than generators [C ] // Proceedings of the 8th International Conference on Learning Representations . 2020 .

SUN Y , WANG S H , LI Y K , et al . ERNIE: enhanced representation through knowledge integration [J ] . arXiv preprint arXiv:1904. 09223 , 2019 .

DEVLIN J , CHANG M W , LEE K , et al . BERT: pre-training of deep bidirectional transformers for language understanding [J ] . arXiv preprint arXiv: 1810.04805 , 2018 .

CUI Y M , CHE W X , LIU T , et al . Revisiting pre-trained models for Chinese natural language processing [J ] . arXiv preprint arXiv: 2004.13922 , 2020 .

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

No data

Related Author

No data

Related Institution

No data

⁰