#dh2016 — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #dh2016, aggregated by home.social.
-
Now up at #DH2024, Maciej Eder, developer of #stylo and co-organizer of #DH2016 in #Krakow, on various distance measures for #Stylometry: "Manhattan, Euclidean and their Siblings. Exploring Exotic Measures of Text Similarities...".
Key idea: Manhattan distance is L1-norm based, Euclidean is L2. But we can vary this parameter for a wide range of values, from 0.1 to 10. Then evaluate accuracy for authorship attribution.
Result: For longer vectors, it pays off to use a value of less than 1!