Fig. 2.

The 3d Mind Model across languages. A A UMAP (McInnes et al., 2018) projection illustrates the semantic similarity between mental states words as proximity in a 2d space (see SI). The closer the two words are to each other, the more similar their vector representation in the 300d fastText embedding. The placement of words reflects the average across all 17 languages we examined. Darker blue points indicate states with more consistent meanings across languages. B Ratings of states on the dimensions of the 3d Mind Model (x-axes) were strongly correlated with scores from a 3d UMAP projection of the word embedding (y-axis). The solid line shows the linear best fit between the ratings and the cross-language average fastText embedding (*** = p < .001). The lighter lines represent the fit of each of the 17 languages