---
license: cc-by-sa-4.0
datasets:
- Wikidepia/openslr_enhanced
language:
- id
- jv
---

# Montreal Forced Aligner (MFA) for Indonesia-Javanese

This repository contains MFA model for Indonesia-Javanese language. This model primarily trained on Javanese ASR dataset (https://www.openslr.org/35/), that are enhanced using DeepFilterNet2 to remove unwanted noise. Lexicon contained in this repository comes from Google's language-resource [Javanese Lexicon](https://github.com/google/language-resources/blob/master/jv/data/lexicon.tsv).

While this model is only trained on Javanese language, you can also use this to align Indonesian speech. You might need to add Indonesian lexicon to the dictionary file.

## Example Usage

To align:

```bash
mfa align --g2p_model_path g2p_jv.zip audio_dir lexicon_jv.dict acoustic_model.zip aligned_dir
```

## Resources
- https://montreal-forced-aligner.readthedocs.io/en/latest/first_steps/alignment_example.html
- https://mfa-models.readthedocs.io/en/latest/

## License

CC-BY-SA-4.0