Japanese kanji network visualisation of relationships by Dea Bankova

These visualisations are fully-interactive graphs of Japanese kanji by their constituent radicals. Kanji can be viewed grouped by school grade when they are taught, as well as by topic. In each case, the kanji are clustered by radicals. On click of each kanji, basic information about it from the online dictionary Jisho (meaning, reading) is displayed.

In practice, Japanese kanji contain one radical each.
I set out to create a modified network graph where each kanji can contain multiple radicals ('building blocks'). My hypothesis was that visualising kanji like this, i.e. using a force graph that naturally nudges kanji which contain similar building blocks together, might help with kanji memorisation - a very challenging task, especially for someone who doesn't have good visual memory, such as myself.

I also used the classification by concept by Wiki dataset to explore patterns between kanji belonging to similar meaning categories and their constituent parts.

It seems like some interesting clear patterns emerge, such the large number of appearances of the radical 心 (heart) in emotion-revealing kanji.

Initially, I also considered using simple machine learning techniques to automatically cluster kanji in meaning groups based on their constituent radicals. This approach didn't yield visually intuitive results.