Search

Computational methods have found a wide area of application in social sciences and humanities. These methods focus on collecting and analysing large quantities of data gleaned from digital environments. Although such methods are interdisciplinary and diverse, they all tend to share three common features: automated or semi-automated collection of data from digital environments, large scale datasets, and the usage of a range of different techniques drawn from computer science to model and analyse the data. Yet, beyond the argued impact in re-defining scientific method (Anderson, 2008 ; Floridi, 2014 ; Kitchin, 2014), problems around the applicability and epistemology of computational methods were soon outlined by some scholars (boyd and Crawford, 2012). A lack of theoretically informed inferences (Coveney, Dougherty, Highfield, 2016 ; Boyd, 2018) is often found in some studies, supported by an assumption that computational methods allow data to speak for itself. Recent scholarship has found a fruitful connection between theoretical elements of the political economy of communication (PEC) and various computational techniques (e.g. Bilić, Furman, Yildirim, 2018 ; Furman, Saka, Yildirim, Elbeyi, 2019 ; Birkinbine and Gomez, 2020). PEC provides a broad conceptualisation of the role of the media in society. It looks at the structural position (i.e. media ownership) and connects it with opportunities, constraints and limits to publishing and framing news content in certain ways. In this presentation, we use PEC to select media outlets based on audience reach, regional coverage, public ownership, and non-profit ownership with media frame analysis that looks at the salience of specific linguistic elements in news reports. This allows us to make theoretically-informed inferences on the thematic scope of news media output in relation to the COVID-19 pandemic. Comparing two peak periods of the COVID-19 pandemic during 2020, the empirical part of our study will apply a computational methodology and prepare a semantic textual analysis of all COVID-19 related articles published by 21 Croatian digital news media outlets. This methodology will use natural language processing and correspondence analysis to explore which news concepts relate to which media, and which news themes and frames dominated two peak periods in relation to different structural positions of the media. The findings of our exploratory study will be presented through visual representations of the complex dataset.