Multi-label Text Classification on EURLEX57K (test)
79.6RP@5 (All Labels)BERT-BASE
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| BERT-BASEdocument tokens=first 512 tokens2019.06 | 79.6 | 82.3 | 73.2 | 83.5 | 84.6 | 68.6 | 63.6 | 2.8 | 2.3 | |
| BIGRU-LWAN-ELMOdocument tokens=first 512 tokens, embeddings=ELMo2019.06 | 78.1 | 81.1 | 71.9 | 82.1 | 83.5 | 66.8 | 61.9 | 4.4 | 2.8 | |
| BIGRU-LWAN-L2Vdocument tokens=full, embeddings=L2V2019.06 | 77.5 | 80.4 | 71.1 | 81.5 | 82.8 | 65.6 | 61.2 | 3.4 | 2.4 | |
| BIGRU-LWAN-L2Vdocument tokens=first 512 tokens, embeddings=L2V2019.06 | 77 | 79.6 | 70.9 | 81.1 | 82.5 | 64.1 | 60 | 4.7 | 3 | |
| BIGRU-LWANdocument tokens=full, embeddings=GloVe2019.06 | 76.6 | 79.6 | 69.8 | 80.5 | 81.9 | 66.2 | 61.8 | 2.9 | 1.9 | |
| BIGRU-ATTdocument tokens=full, embeddings=GloVe2019.06 | 75.8 | 78.9 | 68.9 | 79.9 | 81.3 | 63.1 | 58 | 4 | 2.7 | |
| HANdocument tokens=full, embeddings=GloVe2019.06 | 74.6 | 77.8 | 68 | 78.9 | 80.5 | 59.7 | 54.4 | 5.1 | 3.4 | |
| ZERO-BIGRU-LWANdocument tokens=full, embeddings=GloVe, zero-shot mode=true2019.06 | 71.8 | 75.2 | 65.2 | 76.4 | 78 | 56.1 | 51 | 43.8 | 34.5 | |
| CNN-LWANdocument tokens=full, embeddings=GloVe2019.06 | 71.6 | 74.6 | 64.2 | 76.1 | 77.2 | 61.3 | 55.7 | 3.6 | 2.3 | |
| Logistic Regressiondocument tokens=full, embeddings=GloVe2019.06 | 71 | 74.1 | 53.9 | 76.7 | 78.1 | 50.8 | 47 | 1.1 | 1.1 | |
| ZERO-CNN-LWANdocument tokens=full, embeddings=GloVe, zero-shot mode=true2019.06 | 68.4 | 71.7 | 61.8 | 73 | 74.5 | 49.5 | 45.4 | 32.1 | 26.4 | |
| Exact Matchdocument tokens=full2019.06 | 9.7 | 9.9 | 12 | 21.9 | 20.1 | 11.1 | 7.4 | 19.4 | 18.6 |