Between Article and Topic: News Events as Level of Analysis and Their Computational Identification

Abstract

When comparing media coverage or analysing which content people are exposed to, researchers need to abstract from individual articles. At the same time, aggregating them into broad topics or issues often is too coarse and loses nuance. Both theoretically and methodologically, the analysis of an appropriate intermediate level of aggregation is underdeveloped. This article advances research in various areas of journalism studies by developing a theoretical argument for introducing the "news event" as a level of analysis. Based on this, we discuss several computational approaches to empirically detect such news events in large corpora of news coverage in an unsupervised manner. We provide two approaches: One based on traditional tfÁidf based cosine similarities, and one that relies on word embeddings, in particular the softcosine measure. Both methods, combined with a network clustering algorithm, perform very well in detecting news events. We apply this method in a case study of 45k news articles from different outlets, in which we show that different news outlets have distinct profiles in the events they cover.

Linking event archives to news: a computational method for analyzing the gatekeeping process

“…We build on recent work in the field of communication that uses document similarity calculations to study news flows (Boumans et al, 2018;Mozer et al, 2020;Nicholls, 2019;Trilling & Van Hoof, 2020;Welbers et al, 2018). Events and news articles are transformed to sparse weighted vectors, and whether an event is likely to be covered in a news articles is then measured based on the similarity of these vectors and the time between the event and the news article publication.…”

“…Although these fields seem to have focused little on the specific task of analyzing event selection bias (Hamborg et al, 2019) the task of linking events to news is strongly related to techniques for information retrieval and document clustering. In the field of communication science, scholars have also started adopting these techniques for linking news articles based on similarities in what events are covered (Trilling & Van Hoof, 2020;Welbers et al, 2018). The current study builds on and contributes to these approaches, but with a specialized application for the task of matching event databases to news.…”

“…As a general guideline, a lower-dimensional representation works well for matching documents on topical content, but for more specific matches where specific keywords matter, a BOW approach can be more appropriate (Mozer et al, 2020;Trilling & Van Hoof, 2020). For matching many events to many news articles, keywords pertaining to who, what and where are particularly important, so we used a BOW approach.…”

“…A common recommendation for calculating document similarity is to calculate the cosine similarity of the document vectors (Trilling & Van Hoof, 2020). However, for matching short event descriptions to news articles the regular dot product can be a good alternative, and is used in our example analysis.…”

Recommenders with a Mission

“…This can be done through manual annotation or automated extraction process. Two unsupervised learning approaches for doing this automatically can be found in Nicholls and Bright [38] and Trilling and van Hoof [51]. Once the stories are identified, the Fragmentation score can be defined as the aggregate average distance between all sets of recommendations between all users.…”