Showcases transforms official police reports into multimedia installations that trace the normalization of right-wing violence in Berlin and Brandenburg, Germany. From everyday hate crimes to nationalist symbols on public walls, the project confronts what society often overlooks.

Showcases is a data-driven investigation that visualizes police reports of politically motivated crimes to draw attention to a mounting normalization of xenophobic, trans- and homophobic violence and right-extremism in Germany. It roots in disturbing records of incidents such as: swastikas (Hakenkreuz) appearing on walls of mosques, synagogues and schools; people tearing off the Hidschāb from a 14-year-old and spitting in her hair; cashiers confronted with drunk customers that perform the Hitler salute. These are not isolated events, but daily occurrences that suggest an erosion of democratic norms and a resurgence of far-right ideology in society.

The project draws connection to the German political landscape, where with the growing support for the right-extremist party Alternative für Deutschland (AfD) since 2013, politically motivated crimes linked to far-right ideology have increased significantly (Bundeskriminalamt, 2025). This is taking place in a country whose history, marked by authoritarianism and violence, serves as a stark reminder of the intertwining of nationalist parties in the parliament and violence by like-minded extremists. And yet, in much of the public and political discourse, these tendencies seem to be neglected, while populists use their media coverage for pushing toxic narratives and anti-islamic, anti-woke, anti-semitic and anti-democratic agendas.

How it works

Scrape everything → save it raw → filter by categories.json → translate what is new → the site reads the result.

categories.json          categories (define the dataset) + commentary (views on top of it)
pipeline/
  scrape_rss.py          1. daily: Berlin + Brandenburg RSS feeds → data/police_rss_<month>.csv
  scrape_berlin.py          archive backfill                      → data/berlin_police_results.csv
  scrape_brandenburg.py     archive backfill                      → data/brandenburg_police_results.csv
  data/                     every scraped report, unfiltered
  filter.py              2. all of data/ × categories.json        → static/all_merged.csv
  translate.mjs          3. new reports in the dataset → English  → static/translations/<year>.json
src/                     the SvelteKit site (reads static/all_merged.csv and static/translations/)

The GitHub Action rss.yml runs steps 1–3 every day and commits the result; deploy.yml builds and publishes the site.

Categories

Everything is defined in categories.json:

• categories: each has an id (e.g. kennzeichen) and terms (hakenkreuz, hitlergruß, …). Terms are word stems searched in all scraped reports; a match puts the report in the dataset.
• commentary: same shape, but terms are searched only within the dataset (highlights on the categorical timeline). They never pull a report in.

After editing it, run npm run filter. The filter re-scores every scraped report each time (about 10 seconds), so reports that did not match before are picked up.

Commands

npm install
npm run dev               # the site

pip install -r pipeline/requirements.txt
npm run pipeline          # scrape RSS feeds + filter
npm run filter            # filter only (after editing categories.json)
npm run translate         # translate with DeepL (needs DEEPL_API_KEY in .env)
npm run translate:local   # translate the backlog with a local Ollama model (resumable)

Archive scrapers, full re-scans and translation details: pipeline/README.md.