Topic modeling of news articles

Since quite a few things have changed in the world over the last years, I was wondering if and how these changes also reflect in the news.

The analysis examined Estonian Public Broadcasting articles spanning 2016-2021. Articles underwent translation and lemmatization before topic extraction via non-negative matrix factorization, yielding 15 distinct categories: basketball, culture, crime, economy, education, film, football, foreign politics, healthcare, music, other sports, politics, sports general, tennis, and weather.

The pandemic has clearly left its mark — the healthcare topic skyrocketed at the beginning of 2020 and has since remained on a relatively high level.

The author noted economy coverage increased in March 2020 while sports topics diminished. Seasonal patterns emerged in sports and education; crime and music showed trends; politics and healthcare correlated with current events.