Research Article
Fusing Part-of-Speech Information in Low-Resource Neural Paraphrase Generation
Table 2
Performance of RNN-based models on Quora datasets.
| | Dataset | Model | BLEU | ROUGE-1 | ROUGE-2 | ROUGE-L |
| | Quora10K | base | 15.39 (±0.22) | 38.47 (±0.29) | 20.54 (±0.24) | 37.43 (±0.26) | | add | 16.25 (±0.08) | 39.82 (±0.23) | 21.53 (±0.15) | 38.77 (±0.22) | | cat | 16.06 (±0.04) | 39.51 (±0.09) | 21.27 (±0.09) | 38.46 (±0.09) | | dc | 16.8 (±0.13) | 40.79 (±0.3) | 22.02 (±0.24) | 39.64 (±0.32) |
| | Quora20K | base | 17.8 (±0.16) | 43.52 (±0.21) | 24.08 (±0.17) | 42.12 (±0.19) | | add | 18.34 (±0.09) | 44.42 (±0.13) | 24.73 (±0.12) | 43.01 (±0.12) | | cat | 18.21 (±0.12) | 44.12 (±0.06) | 24.52 (±0.04) | 42.73 (±0.06) | | dc | 18.79 (±0.25) | 45.26 (±0.22) | 25.31 (±0.31) | 43.82 (±0.21) |
| | Quora50K | base | 20.57 (±0.12) | 48.63 (±0.15) | 28.12 (±0.15) | 46.84 (±0.13) | | add | 20.89 (±0.11) | 49.1 (±0.06) | 28.4 (±0.1) | 47.27 (±0.08) | | cat | 20.84 (±0.05) | 48.95 (±0.08) | 28.36 (±0.1) | 47.13 (±0.07) | | dc | 21.01 (±0.1) | 49.61 (±0.12) | 28.82 (±0.13) | 47.81 (±0.08) |
|
|
