code that generates graphs to learn kanjis in different spatial ways.
There is not enough dimensions on a screen to adequately plot all kanjis, so it mostly plots random subsets of the kanji space for intelligibility, and plots them in different ways.
It takes as input a CSV for all kanjis, and outputs 4 kinds of graphs:
- Similarity, places the nodes according to visual similarity and draws edges based on semantic similarity
- Composition, a random subset of the total space based on which kanji is component of which
- Oroots, decomposes each kanji into roots and subroots
- Components, highlights all the kanjis derived from given kanjis
It highlights kanjis based on familiarity and puts an emphasis on the recently added ones (+ random spotlight)