Table 2 Dataset statistics for each subtask
From: Hypert: hypernymy-aware BERT with Hearst pattern exploitation for hypernym discovery
Subtask | Corpus size | # of | |||
|---|---|---|---|---|---|
Vocabulary | Train | Valid | Test | ||
1A(English) | 16G | 218,753 | 1500 | 50 | 1500 |
2A(Medical) | 800M | 93,888 | 500 | 15 | 500 |
2B(Music) | 500M | 69,118 | 500 | 15 | 500 |