KG-to-Text Generation
Knowledge-graph-to-text (KG-to-text) generation aims to generate high-quality texts which are consistent with input graphs.
Description from: JointGT: Graph-Text Joint Representation Learning for Text Generation from Knowledge Graphs
Papers
Showing 1–10 of 22 papers
All datasetsWebNLG 2.0 (Unconstrained)WebNLG 2.0 (Constrained)AGENDAEventNarrativePathQuestionWebQuestionsWikiGraphsWebNLG (All)WebNLG (Seen)WebNLG (Unseen)ENT-DESC
Benchmark Results
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | GAP - Me,re | ROUGE | 76.22 | — | Unverified |
| 2 | GAP - Me,r+γ | BLEU | 66.2 | — | Unverified |
| 3 | JointGT (T5) | BLEU | 66.14 | — | Unverified |
| 4 | JointGT (BART) | BLEU | 65.92 | — | Unverified |
| 5 | JointGT (BART) - w/ JointGTPretrain | BLEU | 65.92 | — | Unverified |
| 6 | JointGT (BART) - w/ BARTPretrain | BLEU | 64.6 | — | Unverified |
| 7 | BART | BLEU | 64.55 | — | Unverified |
| 8 | T5 | BLEU | 64.42 | — | Unverified |
| 9 | KGPT | BLEU | 64.11 | — | Unverified |
| 10 | KGPT w/o pretrain | BLEU | 62.3 | — | Unverified |