Angyalffy, Dávid – Németh, Renáta – Buda, Jakab
Médiakutató, 2026. tél, XXVII. évf., 4. sz.
The quantitative analysis of changes in editorial policy has become a mainstream tool in media research, owing to advances in computational methods. This paper examines changes in the content of Index following the 2020 change in ownership, based on machine learning methods. It employs a proxy-based approach: first, the researchers trained three models, each sensitive to different textual characteristics, on HVG’s and Origo’s domestic news articles from 2019 and 2021, respectively; then, they used the trained models to classify Index’s articles from the same time periods into HVG-like and Origo-like groups. The findings suggest that, in terms of certain linguistic and text-level patterns, Index shares more similarities with the Origo reference corpus. The interpretability analysis also identifies the stylistic and thematic markers of this shift.