Skip to main content

Table 5 Macro-F1 scores on out-of-domain datasets. 95% Confidence intervals in square brackets (computed over 10,000 bootstrap samples)

From: LEIA: Linguistic Embeddings for the Identification of Affect

 

LIWC

NRC

NBSVM

LEIA-base

LEIA-large

Universal Joy

23.45 [19.97,27.29]

28.98 [26.15,31.99]

41.70 [37.36,46.08]

54.18 [48.79,59.88]

54.17 [48.68,59.84]

GoEmotions

45.81 [42.72,48.63]

32.68 [31.18,34.25]

48.23 [45.85,50.59]

46.31 [43.98,48.72]

45.75 [43.45,48.09]

TEC

36.02 [34.02,37.99]

33.92 [32.65,35.27]

39.07 [37.28,40.92]

43.87 [42.05,45.61]

44.12 [42.34,45.89]

SemEval

66.72 [64.61,69.1]

49.86 [48.27,51.4]

68.77 [66.29,71.25]

71.68 [69.18,74.19]

70.04 [67.48,72.52]

enISEAR

23.51 [19.43,26.89]

42.72 [40.26,44.89]

55.33 [51.22,59.41]

70.37 [66.63,74.01]

79.94 [76.69,83.14]