Measuring Attitudes to Migration in the Media automatically with Complementary Data Sources and Methods

Abstract

In this study, we trace trends and the evolution of connotations of migrants and refugees in the US from 1987 to 2018, using automatic methods of text analysis. To do so, we apply methods from computational linguistics research, and discuss their respective added-value. We rely on complementary data sources: newspaper articles, transcripts of TV broadcast and social media data. More specifically, we diachronically measure attitudes to migra- tion, by analysing spikes in media discourse, associations that terms from the domain of migration typically have, and positive and negative stance that the topics in the migration discourse have provoked over the last two decades. Our results show that: (i) elections and important events lead to spikes, and the interest in migration increases over time; (ii) legal issues grow more important in media discourse of migration; (iii) distributional semantics confirms the increase of legal concerns, and demonstrates that migrant and refugee are often used as synonyms; (iv) sentiment towards migration is generally becoming more positive.