Project

Aliases of Arda

Tolkien's characters answer to many names — Aragorn is Strider, Elessar, and Estel; Gandalf is Mithrandir and Olórin. This project asks a concrete question: how much does the modern contextual-embedding stack dissolve the rare-name aliasing problem that an earlier word2vec approach couldn't handle?

The original tool leaned on static word vectors, which struggle when a name appears only a handful of times across the text. Rare names end up with thin, noisy representations — precisely the cases where knowing that two names point to one character matters most. The reboot revisits that weakness with contextual embeddings and measures the difference against a hand-checked set of known alias pairs.

Results

The findings, tables, and figures will live here as the analysis comes together.