用AI分析媒体情绪，预测毛里求斯大选结果。
Leveraging AI and Sentiment Analysis for Forecasting Election Outcomes in Mauritius
- 用多语言BERT和自定义算法分析新闻情绪倾向
- 预测变革联盟至少获37席，联盟莱佩普得23席
- 适合无可靠民调地区的政治预测参考
本研究探索了基于人工智能的情绪分析在预测毛里求斯2024年选举结果中的应用。由于缺乏可靠的民调数据，我们通过分类主流毛里求斯媒体对两大政党——'联盟莱佩普'与'变革联盟'的报道，将其情绪分为正面、负面或中性。采用多语言BERT模型与自定义情绪评分算法量化情绪动态，并引入情绪影响得分（SIS）评估情绪随时间的影响。预测模型显示，'变革联盟'有望至少赢得37个议席，'联盟莱佩普'则预计获得剩余23席，共60席。结果显示，正面媒体报道与预期选举优势高度相关，凸显媒体在塑造公众认知中的作用。该方法通过调整评分缓解了媒体偏见，为缺乏传统民调基础设施的地区提供了一种可靠的替代方案。研究提出了一种可扩展的政治预测方法，推动了政治数据科学的发展。
原文摘要 · Abstract (English)
This study explores the use of AI-driven sentiment analysis as a novel tool for forecasting election outcomes, focusing on Mauritius' 2024 elections. In the absence of reliable polling data, we analyze media sentiment toward two main political parties L'Alliance Lepep and L'Alliance Du Changement by classifying news articles from prominent Mauritian media outlets as positive, negative, or neutral. We employ a multilingual BERT-based model and a custom Sentiment Scoring Algorithm to quantify sentiment dynamics and apply the Sentiment Impact Score (SIS) for measuring sentiment influence over time. Our forecast model suggests L'Alliance Du Changement is likely to secure a minimum of 37 seats, while L'Alliance Lepep is predicted to obtain the remaining 23 seats out of the 60 available. Findings indicate that positive media sentiment strongly correlates with projected electoral gains, underscoring the role of media in shaping public perception. This approach not only mitigates media bias through adjusted scoring but also serves as a reliable alternative to traditional polling. The study offers a scalable methodology for political forecasting in regions with limited polling infrastructure and contributes to advancements in the field of political data science.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。