## Inverse Text Normalization: Weighted Finite State Transducer

### Model Overview

This model is an OpenFST finite state archive (.far) for use within the opensource Sparrowhawk normalization engine [2].  This model has been created for Dutch (nl-BE) text.

### Model Architecture

This model uses weighted finite-state transducer (WFST) grammars that map strings in spoken form to strings in written form.

### Training

The model was constructed using NeMo toolkit as outlined in [Inverse Text Normalization](https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en/main/nlp/text_normalization/intro.html).

### How to Use this Model

The [Riva Quick Start Guide](https://ngc.nvidia.com/catalog/resources/nvidia:riva:riva_quickstart) is recommended as the starting point for trying out Riva models. For more information on using this model with Riva Speech Services, see the [Riva User Guide](https://docs.nvidia.com/deeplearning/riva/user-guide/docs/asr/asr-overview.html).

### Input

This model is part of a pipeline for speech to text and accepts text as input.

### Output

The output is written form text.

### References

[1] [NeMo Inverse Text Normalization: From Development To Production](https://arxiv.org/abs/2104.05055)
[2] [Google Sparrowhawk](https://github.com/google/sparrowhawk)

### License

By downloading and using the models and resources packaged with Riva Conversational AI, you accept the terms of the Riva [license](https://www.nvidia.com/en-us/data-center/products/nvidia-ai-enterprise/eula/).

Riva ASR Dutch (nl-BE) Inverse Normalization Grammar

Inverse Text Normalization: Weighted Finite State Transducer

Model Overview

Model Architecture

Training

How to Use this Model

Input

Output

References

License