GameSpace is an explorable three-dimensional ontological space in which each of our natural language processing model’s games are represented as a data-rich “stars” whose positions are semantically meaningful. Specifically, games are placed in the space such that their most related games are nearby. Three-dimensional coordinates for the games were derived by submitting their LSA vectors to LLE, as in GameGlobs. The user can fly freely through the space using conventional 3D game controls, and upon encountering a game she can engage it (by clicking it) to explore data about it: its title and year of release; an embedded YouTube player with a Let’s Play video preloaded; and an embedded pane displaying its Wikipedia page. Below, the central cluster represents combat-oriented games, with racing games forming a dense cluster to the bottom-right. This is a collaboration with James Ryan at the Expressive Intelligence Studio.

GameSpace is not yet live, expected release is February 2017. For more information please check out publications.