基于财经新闻的金融领域负面情感词典构建研究
Research on the Construction of Negative Sentiment Dictionary in Finance Based on Financial News
赵又霖 1林怡妮 2孙虹 2程丽洁 2徐竟楠 2陆颖隽3
作者信息
- 1. 南京大学信息管理学院,江苏南京 210023;河海大学商学院,江苏南京 211100
- 2. 河海大学商学院,江苏南京 211100
- 3. 武汉大学信息管理学院,湖北武汉 430072
- 折叠
摘要
金融领域的情感倾向具有其领域特有性,通用情感词典无法准确判断其情感倾向.本文以道琼斯工业平均指数排名前30的公司的实时新闻信息为研究对象,设置人工标注规则,采用多种机器学习算法对已标注数据集进行训练,使用特征选择算法抽取特征词语作为负面情感词语,从而构建金融领域负面情感词典.本文以McDonald Financial Dictionary(麦当劳金融词典)为目标情感词典,通过直接和间接评测两种方式验证其准确性.研究结果表明:本文通过实验获得金融领域603个负面的领域情感词,其领域负面情感词典的覆盖率为82.5％,且对金融领域新闻信息情感识别的准确率达到92.8％,能够显著提高金融领域负面情感倾向分析的准确率.
Abstract
The emotional tendencies in the field of economics have their own domain-specific characteristics,and general sentiment dictionaries cannot accurately judge their emotional tendencies.This paper takes the real-time news information of the top 30 companies in the Dow Jones Industrial Average as the research object,sets up manual labeling rules,uses a variety of machine learning algorithms to train the labeled data sets,and uses the feature selection algorithm to extract thie characteristic words as negative emotion words,so as to construct the negative emotion dictionary in the field of finance.Taking the current classic McDonald Financial Dictionary as the target sentiment dictionary,the accuracy of the domain dictionary constructed based on dynamic news information in this study is verified and evaluated by direct and indirect evaluation.603 words have been successfully gotten in this research,and the results show that the coverage rate of the negative sentiment dictionary in the financial field constructed in this paper is 82.5％,and the accuracy of the recognition of financial news sentiment reaches 92.8％.The financial negative sentiment dictionary constructed in this paper can significantly improve the accuracy of negative sentiment tendency in finance.
关键词
负面情感词典/财经新闻/特征词语/直接评测/间接评测Key words
Negative emotion dictionary/Financial news/Characteristic words/Direct evaluation/Indirect evaluation引用本文复制引用
基金项目
江苏省社科基金青年项目(20TQC001)
中国博士后科学基金特别资助(2021T140311)
中国博士后科学基金面上项目(2019M650108)
中央高校基本科研业务费项目(B230207061)
出版年
2024
NETL
NSTL
万方数据