Table 4 Macro-F1 scores on the Vent test sets. 95% Confidence interval in square brackets (computed over 10,000 bootstrap samples). For LIWC and NRC, we only consider 4 out of 5 labels and perform binary classification for each label using the “one-vs-rest” setting
From: LEIA: Linguistic Embeddings for the Identification of Affect
LIWC | NRC | NBSVM | LEIA-base | LEIA-large | |
|---|---|---|---|---|---|
User | 32.88 [32.79,32.96] | 32.91 [32.85,32.97] | 60.15 [60.05,60.25] | 72.92 [72.82,73.02] | 73.37 [73.28,73.46] |
Temporal | 34.64 [34.55,34.72] | 33.07 [33.01,33.12] | 60.52 [60.42,60.63] | 73.03 [72.95,73.11] | 73.43 [73.34,73.53] |
Random | 32.84 [32.74,32.94] | 33.02 [32.95,33.08] | 60.26 [60.16,60.36] | 73.02 [72.94,73.12] | 73.57 [73.48,73.66] |