Use the NGC CLI to download:
Copied!
Model
English single-pass tagger-based model for inverse text normalization based on bert-base-uncased, trained on 2 mln sentences from Google Text Normalization Dataset, achieves 3.75% WER on Google default test set
Use the NGC CLI to download:
Copied!
1 Version
1.9.0Selected05/12/2022 3:37 PM UTC424.78 MB5 EpochsBatch Size: 128GPU: V100Copied!
1.9.0Selected
05/12/2022 3:37 PM UTC424.78 MB5 EpochsBatch Size: 128GPU: V100
Copied!
Accuracy
| Key | Value |
|---|---|
| default test (from Google Text Normalization Dataset) | 3.75% WER, 97.47% Sentence Accuracy |
Model
| Key | Value |
|---|---|
| Architecture | BERT |
| Outputs | List of tab-separated text records, consisting of 5 columns, first column is the final text after ITN |
| Inputs | List of spoken-domain sentences without punctuation, as in ASR output |