摘要
本文旨在探讨ChatGPT类大语言模型(以下简称ChatGPT)在舆情精准研判场景中的基础理论和实证应用。在理论部分,提出“大数据、大模型、大计算”全新范式驱动的舆情精准研判。深入剖析了ChatGPT的工作原理,阐述了它与传统舆情分析方法相比具备的优势和存在的局限。在对ChatGPT的特点和舆情精准研判的要义进行厘定的基础上,对舆情研判中人工智能(AI)的应用现状加以分析。随后,从学理方面探析了如何利用大语言模型进行复杂舆情的精准研判以及ChatGPT在舆情精准研判中的实际应用,分析其在文本分析、情绪识别、舆情预测和预警中扮演的角色。理论探讨的同时取用了认知科学和传播学的视角,为舆情精准分析和研判提供了一个全新的理论框架。实证部分,利用ChatGPT大模型预测热门话题和事件是否会登上热搜(以“登上热搜”作为重大舆情事件和事件舆情的重要表征),并进行了大量的实证研究。研究结果表明,ChatGPT类大语言模型具有较高的预测精度。它能在网络早期流传的信息中,洞察潜在的网络热点,提前判定可能引发社会关注的事件,从而具备预警功能。这项实证研究,有力地阐明了ChatGPT类大语言模型在实际应用中的显著效用,并与文中理论探讨相呼应,揭示了大语言模型类先进AI技术和“大数据、大模型、大计算”全新范式在舆情精准研判中的重要地位、价值和作用。
This article aims to explore the basic theories and empirical applications of ChatGPT-like large language models(hereinafter referred to as ChatGPT) in the context of precise public opinion analysis.In the theoretical section,it introduces a new paradigm driven by "big data,big models,big computation" for precise public opinion judgment.It delves into the working principles of ChatGPT,outlining its advantages and limitations compared to traditional public opinion analysis methods.Upon establishing an understanding of ChatGPT's characteristics and the essentials of precise public opinion analysis,the article analyzes the current state of artificial intelligence(AI)application in public opinion judgment.It then theoretically explores how to utilize large language models for the precise judgment of complex public sentiments and the practical application of ChatGPT in this field,analyzing its role in text analysis,emotion recognition,and forecasting and early warning of public opinion trends.The theoretical discussion also adopts perspectives from cognitive science and communication studies,providing a new theoretical framework for precise analysis and judgment of public opinion.In the empirical section,the article utilizes the ChatGPT model to predict whether hot topics and events will trend on social media(using "trending" as a significant indicator of major public opinion events and their sentiment),conducting extensive empirical research.The results show that ChatGPT-like large language models have a high predictive accuracy.They can identify potential online hotspots in the early stages of information dissemination on the internet,determining in advance the events that may attract social attention,thus serving a warning function.This empirical study effectively demonstrates the significant utility of ChatGPTlike large language models in practical applications and resonates with the theoretical discussions in the article,revealing the important position,value,and role of advanced AI technologies like large langua
作者
丁晓蔚
赵鑫
刘梓航
刘天昊
Ding Xiaowei;Zhao Xin;Liu Zihang;Liu Tianhao
出处
《传媒观察》
CSSCI
2024年第3期45-61,共17页
Media Observer
基金
南京大学文科青年跨学科团队专项“中央高校基本科研业务费专项资金资助项目”(010814370112)
南京大学文科卓越中长期研究专项“南京大学新时代文科卓越研究计划中长期研究专项资助项目”(010814914215)的研究成果。
