Posts
Newest first-
06
Names or story? Splitting every Marvel lookalike into what it shares
A cosine similarity is a sum over words, so we can say exactly how much of every match comes from shared names. For 98% of pages the TF-IDF lookalike is mostly a name: Storm matches the Human Torch on one word. Take the names away and three quarters of the lookalikes change. The new ones are symbiotes, vampires and powered armor, often between pages Wikipedia never linked.
→ -
05
Wikipedia copies the paperwork, not the plot
The Marvel pages get their text. We hunt for passages that appear word for word on two pages: they are publication histories, creator anecdotes and reading lists, almost never the fictional biographies. Every copied paragraph sits between pages that are already linked.
→ -
04
Where did the Renaissance go? 200 Louvain runs and the philosophers no tradition will keep
We switch to the philosopher network and run Louvain 200 times. Almost everyone keeps their tradition, except Galileo, Kepler, Bruno and the other Renaissance naturalists: a tight bloc that a different tradition absorbs each run.
→ -
03
Attack vs. random failure: how many heroes does it take to break the Marvel web?
Removing characters by betweenness, by degree, and at random, and watching the giant component shrink. Plus which single character fragments the network the most, and whether betweenness would pick the same one.
→ -
02
Degree distribution vs. random models: heavy-tailed, BA-shaped, and not a clean power law
Comparing the real Marvel graph against a matched random graph and a tuned Barabási–Albert model, then testing whether a strict power law is statistically supported.
→ -
01
Meet the network: 303 heroes, 17 loners, and one island of Morituri
Loading the frozen week-1 snapshot, reconciling directed and undirected edge counts, mapping who's most linked-to and who links out the most, and finding a nine-node clique that Wikipedia forgot to connect to the rest of the Marvel universe.
→