Prerequisite Chain Learning on NLP to BIO domain
80.6F1 ScoreCLS + Node2vec
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| CLS + Node2vecBase Embedding Model=Node2vec, Learning Protocol=Supervised / Upper Bound2021.05 | 80.6 | 79.56 | 75.47 | 87.27 | |
| CD-VGAE + P2VBase Embedding Model=P2V, Learning Protocol=Zero-shot / Proposed2021.05 | 65.12 | 65.91 | 66.67 | 63.64 | |
| CD-VGAE + BERTBase Embedding Model=BERT, Learning Protocol=Zero-shot / Proposed2021.05 | 62.89 | 62.73 | 64.25 | 63.64 | |
| VGAE + P2VBase Embedding Model=P2V, Learning Protocol=Zero-shot / Baseline2021.05 | 61.77 | 62.73 | 65.21 | 60.91 | |
| VGAE + BERTBase Embedding Model=BERT, Learning Protocol=Zero-shot / Baseline2021.05 | 60.11 | 60.91 | 61.85 | 59.09 | |
| CLS + BERTBase Embedding Model=BERT, Learning Protocol=Zero-shot / Baseline2021.05 | 39.3 | 60 | 74.81 | 27.27 | |
| CLS + P2VBase Embedding Model=P2V, Learning Protocol=Zero-shot / Baseline2021.05 | 22.22 | 53.33 | 60 | 13.64 |