The semantic vocabulary of the Estonian Varandusemuistendid corpus: 36 categories of words (locations, body parts, building types, treasure-wealth terms, mythological beings…), grouped into 6 hierarchical clusters, plus 31 emotion domains. 34,503 wordforms have been classified by an ensemble of 15 annotation methods — 6 semantic (GLiNER, WordNet, NRC, dictionaries, thesaurus, morphological) and 9 emotion (GoEmotions, SetFit, NRC, Vainik lexicon, EKKD-RV, and more).
PMI (Pointwise Mutual Information) matrix showing how strongly category pairs co-occur across 93 treasure types. Blue = strong positive co-occurrence, red = negative (anti-correlated). Click a cell to compare categories.