That sounds fascinating—an opportunity to turn history into a network graph and then apply clustering or community detection. We could start by digitizing old itineraries, geolocating stops, and assigning timestamps. Once we have a clean dataset we can plot the routes, compute centrality metrics, and look for patterns like hub towns or seasonal variations. If we overlay trade volumes or commodity types, maybe we'll uncover hidden trade corridors or see how political events shifted routes. Let's gather the data and see what the numbers reveal.

Sounds like a solid plan. For starting data I’d recommend a few sources that already digitize historical itineraries: the David Rumsey Map Collection has a good selection of Silk Road maps, the Library of Congress has spice trade charts, and the World Digital Library has some medieval caravan routes. For extracting coordinates, tools like GeoGuessr or even the Georeferencer extension for QGIS can help align the old maps to modern coordinates. Once you have the points, we can load them into a GeoPandas dataframe, then use NetworkX to build the route graph. From there, calculating betweenness or PageRank will give us the centrality heat‑map you mentioned. Let’s pull a few of those maps, georeference them, and see what the data says.

Let’s start with a clean, well‑labelled Silk Road map from the David Rumsey collection. It already has latitude and longitude annotations, so we can grab the shapefile or the CSV of points directly. That will save us the georeferencing step for now. Once we have the point data in GeoPandas, I’ll write a short script to connect consecutive stops and calculate betweenness. We can then add the spice lanes as a second layer to compare. Sound good?