Fig. 4: Accuracy obtained by the Pl@ntBERT-large-species model on different typologies (results averaged over the ten cross-validation folds). | Nature Plants

Fig. 4: Accuracy obtained by the Pl@ntBERT-large-species model on different typologies (results averaged over the ten cross-validation folds).

From: Learning the syntax of plant assemblages

Fig. 4

a, Accuracy results (in %, with the best accuracy in green and the worst accuracy in red) of Pl@ntBERT across the European Red List of Habitats categories (DD, data deficient; LC, least concern; NT, near threatened; VU, vulnerable; EN, endangered; CR, critical endangered). b, Accuracy results (in %, with the best accuracy in green and the worst accuracy in red) of Pl@ntBERT across level 1 (that is, broad habitat groups) of the EUNIS typology (MA2, marine; N, coastal; Q, wetlands; R, grasslands; S, heathlands; T, forests; U, inland; V, man-made). c, Comparison of the accuracy (in %, best accuracy in bold) of the models across grouped European Red List of Habitats statuses (data deficient, {DD}; not threatened, {LC + NT}; threatened, {VU + EN + CR}) and EUNIS broad habitat groups (sparse, {MA2 + U}; vegetated, {R + N + Q + S + T}; anthropogenic, {V}).

Back to article page