Table 14 Wilcoxon signed-rank test results of the proposed method against comparison models for the 2B dataset with 10 iterations
From: Hypert: hypernymy-aware BERT with Hearst pattern exploitation for hypernym discovery
Comparison methods | Evaluation measures | Proposed versus (p-value) | |
|---|---|---|---|
RMM [11] | MRR | Win | (1.95e–3) |
MAP | Win | (1.95e–3) | |
P@1 | Win | (1.95e–3) | |
P@3 | Win | (1.95e–3) | |
P@5 | Win | (1.95e–3) | |
P@15 | Win | (1.95e–3) | |
SPON [12] | MRR | Win | (1.95e–3) |
MAP | Win | (1.95e–3) | |
P@1 | Win | (1.95e–3) | |
P@3 | Win | (1.95e–3) | |
P@5 | Win | (1.95e–3) | |
P@15 | Win | (1.95e–3) | |
Prompting BERT (is a) [14] | MRR | Win | (1.95e–3) |
MAP | Win | (1.95e–3) | |
P@1 | Win | (1.95e–3) | |
P@3 | Win | (1.95e–3) | |
P@5 | Win | (1.95e–3) | |
P@15 | Win | (1.95e–3) | |
Prompting BERT (such as) [14] | MRR | Win | (1.95e–3) |
MAP | Win | (1.95e–3) | |
P@1 | Win | (1.95e–3) | |
P@3 | Win | (1.95e–3) | |
P@5 | Win | (1.95e–3) | |
P@15 | Win | (1.95e–3) | |