MIT Press

Table 8

CoNLL04 evaluation of the plausability of explanations, which measures the overlap between machine explanations and human annotations. For each method, we pick the higher F1 score between the two human annotators.

Approach	Precision	Recall	F1
Attention	61.06	30.30	38.94
Saliency Mapping	18.79	39.39	24.43
LIME	22.14	53.33	30.09
Unsupervised Rationale	5.35	74.55	9.31
SHAP	18.18	36.36	23.27
CXPlain	21.21	44.55	27.82
Greedy Adding	33.33	38.03	32.21
Our Approach	65.15	59.24	58.97

Approach	Precision	Recall	F1
Attention	61.06	30.30	38.94
Saliency Mapping	18.79	39.39	24.43
LIME	22.14	53.33	30.09
Unsupervised Rationale	5.35	74.55	9.31
SHAP	18.18	36.36	23.27
CXPlain	21.21	44.55	27.82
Greedy Adding	33.33	38.03	32.21
Our Approach	65.15	59.24	58.97

View Large

This Feature Is Available To Subscribers Only