Replication Data for: Measuring Political Narratives in African News Media: A Word Embeddings Approach

Recent advances in text-as-data provide new opportunities to document how elites shape public discourse on contentious issues. Using a novel word embeddings approach to measure elite-driven political narratives in local news, we analyze Kenyan newspaper coverage of the International Criminal Court's (ICC) prosecutions against domestic leaders. We train our embeddings on an original corpus of 5,292 Kenyan newspaper articles from 2007-2020 and identify significant changes in how local media vilifies the ICC before, during, and after major investigations. We find that as the case progressed, the ICC became more strongly associated with terms of bias and illegitimacy, an association which quickly dissipated after the Court terminated its last proceeding. Our approach illustrates the utility of text-based measures of political sentiment on contentious issues, with implications for research on media narratives, the effects of controversial jurisprudence on public discourse, and backlash against international institutions.

文本数据（text-as-data）研究的新近进展，为刻画精英如何塑造争议性议题上的公共话语提供了全新契机。我们采用一种新颖的词嵌入（word embeddings）方法，用以量化本地新闻中精英主导的政治叙事，并分析肯尼亚报纸对国际刑事法院（International Criminal Court, ICC）起诉本国领导人的报道情况。我们基于2007至2020年间的5292篇肯尼亚报纸文章自建语料库训练词嵌入模型，并识别出本地媒体在重大调查的前、中、后期对国际刑事法院的诋毁性表述所发生的显著变化。研究发现，随着案件推进，国际刑事法院与带有偏见和非法性的表述的关联度持续增强；而在该法院终结最后一项司法程序后，这一关联迅速消散。本研究方法展现了基于文本的争议议题政治情感测量手段的应用价值，其研究结论可为媒体叙事、争议性司法实践对公共话语的影响，以及针对国际机构的反弹等领域的研究提供启示。