WSJ
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Speech Recognition | WSJ (92-eval) | WER1.3 | 131 | |
| Dependency Parsing | WSJ (test) | UAS95.74 | 67 | |
| Speech Recognition | WSJ nov93 (dev) | WER3 | 52 | |
| Part-of-Speech Tagging | WSJ (test) | Accuracy97.78 | 51 | |
| ASR Rescoring | WSJ (test) | WER1.2 | 35 | |
| Speech Recognition | WSJ nov92 (test) | WER1.42 | 34 | |
| Unsupervised constituency parsing | WSJ (test) | Max F184.3 | 29 | |
| Constituency Parsing | WSJ Penn Treebank (test) | F1 Score95.84 | 27 | |
| Dependency Parsing | WSJ 10 or fewer words (test) | UAS79.9 | 25 | |
| Automatic Speech Recognition | 80-hour WSJ (dev93) | WER5.7 | 16 | |
| Unsupervised Dependency Parsing | WSJ section 23 (all lengths) (test) | Directed Dependency Accuracy (DDA)65.8 | 16 | |
| Unsupervised Dependency Parsing | WSJ section 23 length <= 10 (test) | DDA77.2 | 16 | |
| Speech enhancement | WSJ0 UNI | PESQ3.15 | 15 | |
| Monaural Speech Separation | WSJ0-2mix | ΔSI-SDR (dB)24 | 13 | |
| Unsupervised POS tagging | WSJ entire corpus (full) | M1 Score80.8 | 13 | |
| Sentence ordering | WSJ (test) | PRA98.38 | 13 | |
| Speech Recognition | WSJ 93 (test) | WER4.98 | 13 | |
| Keyword Spotting | WSJ (test) | AP0.8094 | 12 | |
| Automatic Speech Recognition | WSJ (test) | WER0.01 | 12 | |
| Unsupervised Parsing | WSJ (test) | F1 Score84.3 | 11 | |
| POS tagging | WSJ (dev) | Accuracy97.37 | 11 | |
| Dependency Parsing | WSJ section 23 (test) | UAS96.2 | 10 | |
| Document Coherence | WSJ permuted document (test) | Accuracy98.59 | 8 | |
| Unsupervised Constituency Parsing | WSJ word-level gold trees (test) | F154.08 | 8 | |
| Dependency Tree Compatibility | WSJ Penn Treebank (test) | Compatibility (%) - All0.7274 | 7 |
Showing 25 of 35 rows