An ego network is the neighbourhood of a single word: the terms that occur near it often enough to count as its company, and the links among those terms in turn. Reading one tells you what a word travels with in a particular text — which is not the same as what it means, but is often the more useful question of a corpus.
This explorer builds those networks from classic texts using corpus-internal statistics only: skip-gram co-occurrence windows and network backbone extraction, with no neural models and no external training data. Type a seed word, click a node to expand it, and trace how meaning propagates outward. In compare mode two sources are placed side by side, so the same seed word can be read against two different authors.
Built with kenon, the lab’s open-source package for semantic and co-occurrence networks (source, documentation).