Table-to-text generation on WIKIBIO (test)
44.89BLEU-4Field-gating Seq2seq + dual attention
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Field-gating Seq2seq + dual attentionArchitecture=structure-aware seq2seq, Attention=dual attention2017.11 | 44.89 | 41.21 | |
| Field-gating Seq2seq + dual attention (beam search)Architecture=structure-aware seq2seq, Attention=dual attention, beam size=52017.11 | 44.71 | 41.65 | |
| Field-gating Seq2seqArchitecture=structure-aware seq2seq, Encoder=field-gating encoder2017.11 | 43.74 | 40.53 | |
| Seq2seq + field + pos (concate)Input features=field and position (concate), Attention=word level attention2017.11 | 43.65 | 40.32 | |
| Seq2seq + field (concate)Input features=field (concate), Attention=word level attention2017.11 | 43.34 | 39.84 | |
| Seq2seqAttention=word level attention2017.11 | 42.06 | 38.06 | |
| Table NLMMethod type=statistical language model2017.11 | 34.7 | 25.8 | |
| Template KNMethod type=statistical language model2017.11 | 19.8 | 10.7 | |
| NLMMethod type=statistical language model2017.11 | 4.17 | 1.48 | |
| KNMethod type=statistical language model2017.11 | 2.21 | 0.38 |