检索规则说明:AND代表“并且”;OR代表“或者”;NOT代表“不包含”;(注意必须大写,运算符两边需空一格)
检 索 范 例 :范例一: (K=图书馆学 OR K=情报学) AND A=范并思 范例二:J=计算机应用与软件 AND (U=C++ OR U=Basic) NOT M=Visual
作 者:陈黎明[1] 苏杨 王小燕[1] 冮建伟 张智 CHEN Liming;SU Yang;WANG Xiaoyan;GANG Jianwei;ZHANG Zhi(School of Finance and Statistics,Hunan University,Changsha 410006,China;School of Economics and Trade,Hunan University,Changsha 410006,China)
机构地区:[1]湖南大学金融与统计学院,长沙410006 [2]湖南大学经济与贸易学院,长沙410006
出 处:《计量经济学报》2025年第1期267-292,共26页China Journal of Econometrics
基 金:国家社会科学基金(23ATJ007);长沙市自然科学基金(kq2208048);湖南省研究生科研创新项目(CX20230418)。
摘 要:增进民生福祉是发展的根本目的,也是衡量发展成效的重要标尺.本文利用大数据和以ChatGPT为代表的大语言模型技术,探索在全国范围内准确核算居民主观幸福感的可能性,提出一种基于微博大数据的居民主观幸福指数构建方法,并基于ChatGLM3开发一种针对于社交媒体文本的情感分析大语言模型SentiGLM,以提升微博文本情感分类的准确性和有效性. SentiGLM模型通过在精心设计的多任务指令微调数据集上进行低秩适应训练,提升其在微博文本情感分析任务中的表现.基于约6千万条的微博文本数据,首次在年、月、周、日四种时间粒度上核算中国各地区以及国家层面的主观幸福感指数.研究发现:SentiGLM在微博文本情感分析中的预测效果明显优于传统机器学习模型(如BERT、LSTM和SnowNLP);同时,基于大语言模型的测度方法相比传统问卷调查,展现出更优的经济性、时效性和稳健性,并且在时间和空间上具备更细的颗粒度.Improving residents' happiness is the fundamental purpose of development and an important measure of development effectiveness.Leveraging big data and large language model technologies represented by ChatGPT,this study explores the possibility of accurately measuring residents'subjective happiness on a national scale.We propose a method for constructing a residents'subjective happiness index based on Weibo data and develop a large language model for sentiment analysis,SentiGLM,built on ChatGLM3,to enhance the accuracy and effectiveness of sentiment classification in Weibo texts.The SentiGLM model significantly improves the performance of sentiment analysis tasks on Weibo texts through low-rank adaptation fine-tuning on a multi-task instruction dataset.Based on approximately 60 million Weibo text data points,this study calculates the subjective happiness index at regional and national levels in China for the first time across four temporal granularities:yearly,monthly,weekly,and daily.The study finds that SentiGLM significantly outperforms traditional machine learning models(such as BERT,LSTM,and SnowNLP)in sentiment analysis of Weibo texts.Moreover,compared to traditional survey methods,the measurement approach based on large language models demonstrates superior cost-effectiveness,timeliness,and robustness,while also providing finer granularity in both temporal and spatial dimensions.
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在链接到云南高校图书馆文献保障联盟下载...
云南高校图书馆联盟文献共享服务平台 版权所有©
您的IP:216.73.216.38