Table 1 ROUGE F1 scores for models on the CNN/Daily Mail test set
From: Multi-task learning for abstractive text summarization with key information guide network
Model | ROUGE-1 | ROUGE-2 | ROUGE-L |
|---|---|---|---|
Seq2Seq + attention (150k vocab) | 30.49 | 11.17 | 28.08 |
Seq2Seq + attention (50k vocab) | 31.33 | 11.81 | 28.83 |
Graph-attention | 38.1 | 13.9 | 34.0 |
Hierarchical attention networks | 35.46 | 13.30 | 32.65 |
Seq2Seq with pointer mechanism | 36.44 | 15.66 | 33.42 |
Key information guide network | 37.76 | 16.56 | 34.49 |
KIGN+prediction-guide | 38.95 | 17.12 | 35.68 |
Our model (joint training) | 39.15 | 17.34 | 35.92 |
Our model (given keywords and key sentences) | 40.34 | 17.70 | 36.57 |