---
language:
- ht
thumbnail: null
tags:
- wav2vec2
license: cc-by-nc-sa-4.0
---

# wav2vec2-HAT-0.2K-ALH-base

This repository provides access to a wav2vec2 model for Haitian Creole (hat). It was trained on a dataset consisting solely of fieldwork recordings.

## Model and data description

The model is a *base* model trained on data of the *Atlas linguistique d'Haiti*. The recordings were made available on the [*Cocoon* platform](https://cocoon.huma-num.fr/exist/crdo/meta/cocoon-8ea988d2-bf16-303d-81a0-0c55cc035240). The formatted version of the data set is to be found [here](https://gin.g-node.org/CREAM/Atlas-linguistique-Haiti). Note that the latter link may only accessible to members of the [LLL](https://lll.cnrs.fr/)/CREAM team.

## Intended uses & limitations

This pre-trained wav2vec2 model is distributed under the [Creative Commons Attribution Non Commercial Share Alike 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/) license.

## Fine-tune with Fairseq for ASR with CTC

As this wav2vec2 model was trained with Fairseq, different tools can be used to fine-tune the model for ASR with CTC. The original Fairseq model `checkpoint_best.pt` is also provided to allow further pre-training or fine-tuning with this framework.

## Fairseq to HuggingFace conversion

Conversion from a Fairseq-based model to a HuggingFace-based one was performed using the [following script](https://github.com/LLL-Orleans/convert_wav2vec2_to_hf).

## Acknowledgments

The authors acknowledge the support of the French Agence Nationale de la Recherche (ANR), under grant ANR-20-CE38-0006 (project CREAM). Experiments presented in this paper were carried out using the Grid'5000 experimental testbed, being developed under the INRIA ALADDIN development action with support from CNRS, RENATER and several Universities as well as other funding bodies (see https://www.grid5000.fr). The authors also benefitted from the use of the CaSciModOT (https://cascimodot.fr/) cluster at the Centre de Calcul Scientifique en région Centre-Val de Loire. 

## Referencing this model

```bibtex
@inproceedings{havard-etal-2024-technologies,
  TITLE = {{Technologies de la parole et donn{\'e}es de terrain : le cas du cr{\'e}ole ha{\"i}tien}},
  AUTHOR = {N. Havard, William and Govain, Renauld and Gon{\c c}alves Teixeira, Daphne and Lecouteux, Benjamin and Schang, Emmanuel},
  URL = {https://inria.hal.science/hal-04623051},
  BOOKTITLE = {{35{\`e}mes Journ{\'e}es d'{\'E}tudes sur la Parole (JEP 2024) 31{\`e}me Conf{\'e}rence sur le Traitement Automatique des Langues Naturelles (TALN 2024) 26{\`e}me Rencontre des {\'E}tudiants Chercheurs en Informatique pour le Traitement Automatique des Langues (RECITAL 2024)}},
  ADDRESS = {Toulouse, France},
  EDITOR = {BALAGUER and Mathieu and BENDAHMAN and Nihed and HO-DAC and Lydia-Mai and MAUCLAIR and Julie and MORENO and Jose G and PINQUIER and Julien},
  PUBLISHER = {{ATALA \& AFPC}},
  VOLUME = {1 : articles longs et prises de position},
  PAGES = {686-694},
  YEAR = {2024},
  MONTH = Jul,
  KEYWORDS = {cr{\'e}ole ha{\"i}tien ; enregistrement de terrain ; mod{\`e}les auto-supervis{\'e}s ; reconnaissance de la parole},
  PDF = {https://inria.hal.science/hal-04623051/file/2789.pdf},
  HAL_ID = {hal-04623051},
  HAL_VERSION = {v1},
}
```