Legitimacy concern, or everyday work? Using Word Embedding Models to Map Representation of International Organizations in News Media

Coverage of international organizations (IOs) in news media is essential for their image in the eyes of general public and, ultimately, for the long-term build-up of their legitimacy. This article sheds new light on debates about the legitimacy and contestation of IOs by estimating the frequency of legitimacy-related claims towards IOs in media across the world. It further classifies the news referring to IOs based on whether they relate to central IO politics, ‘boots-on-the-ground’ policy/project implementation by IOs, or IOs’ unique expertise. By doing so, the article highlights the essential role of everyday, non-reflective and to a large extent non-political coverage of IOs in media, as compared to the rare instances in which a reflective critical or positive stance towards IOs is present in media. Empirical data used consist of a sample from approximately 4 million news articles from 150 countries in the years 2018-2021, selected based on audience geography data. The coverage of four major IOs is analyzed: United Nations, World Bank, World Health Organization, and World Trade Organization. Non-English news (around 80% of analysed content) underwent automated translation. The analysis is rooted in a deep learning word embedding model based on the pre-trained Bidirectional Encoder Representations from Transformers (BERT) semantic model and its infrastructure, combined with the SentenceTransformers model (Sentence-BERT). It is further supported with a detailed hand-coding of a representative sample of 2000 news articles.