ChatGPTOpen DataGeocodingJavascriptChart.js

Mapping Berlin Wall Pieces Around the World with ChatGPT Assistance

This project started after seeing six Berlin Wall remnants on Dalhousie's Agricultural Campus in Truro, Nova Scotia. I wondered how many other pieces had ended up outside Germany.

With ChatGPT assistance, a Wikipedia list became a static site with a world map, searchable table, dashboard, and an explanation of what the project actually counts.

Turning a list into location records

The source is Wikipedia's List of Berlin Wall segments. To make it useful for a map, the project normalizes entries into one row per named city location where possible, then stores the results as CSV and JSON.

That is more than a formatting change. Some source entries group several sites. Other entries mention a piece without saying how many objects are displayed. The data needs to preserve those distinctions.

What counts as a piece

The site uses the broader term “wall pieces” because the source mixes segments, sections, remnants, and fragments. Counting them does not mean every object is a comparable full concrete panel.

Where the source gives a clear number, the project uses it. Where a location is described as having a piece without a larger stated count, the record uses a documented default of one. Entries split from grouped descriptions keep a marker for that treatment when the location-level count remains uncertain.

Making those rules explicit is more useful than presenting a single total as an official inventory.

Showing the same data in several ways

Leaflet provides the map, with marker sizes reflecting the listed number of pieces at each location. A table supports searching and sorting, while Chart.js summarizes the collection by geography and count distribution.

Geocoding is practical rather than authoritative. Some coordinates refer to an institution or city when the exact installation point is not established by the source.

Where AI helped

The project README credits ChatGPT with initial data structuring and static-site generation. I retained responsibility for the idea, review, and publication.

The finished site is a conventional static data application. It does not need a live model to display the map or filter the records.

That makes it a useful example of AI-assisted development and analysis: the model helps produce a durable artifact, while the source definitions and counting rules remain visible for readers to inspect.

Explore the project