ContainersNVIDIADeepSeek-R1NVIDIALlama-3.1-Nemotron-70B-InstructNVIDIAPyTorch
CollectionsNVIDIAOmniverse Kit (FB)NVIDIADeepStream SDKNVIDIAOmniverse Kit App Streaming
ModelsNVIDIAStyleGAN3 pretrained modelsNVIDIAPeopleNetNVIDIATrafficCamNet
ResourcesNVIDIARiva Skills Quick Start
Helm ChartsNVIDIAGPU OperatorNVIDIANVIDIA NIM Operator
NVIDIA
NVIDIA
Russian Tagger-based Inverse Text Normalization
Model
NVIDIA
NVIDIA
Russian Tagger-based Inverse Text Normalization

Russian single-pass tagger-based model for inverse text normalization based on BERT encoder, trained on 2 mln sentences from Google Text Normalization Dataset, achieves 3.55% WER on Google default test set

1 Version
1.11.0Selected
07/21/2022 8:27 PM UTC695.47 MB
Accuracy
KeyValue
DEFAULT TEST (FROM GOOGLE TEXT NORMALIZATION DATASET)3.55% WER, 92.96% SENTENCE ACCURACY
Model
KeyValue
ARCHITECTUREBERT
INPUTSLIST OF SPOKEN-DOMAIN SENTENCES WITHOUT PUNCTUATION, AS IN ASR OUTPUT
OUTPUTSLIST OF TAB-SEPARATED TEXT RECORDS, CONSISTING OF 5 COLUMNS, FIRST COLUMN IS THE FINAL WRITTEN-DOMAIN SENTENCE