asr-puno-quechua
Collection
Automatic speech recognition for Puno Quechua (qxp) using Mozilla Common Voice data. • 7 items • Updated
wav2vec2-base-qxp-finetuned
Model Description
This model is a fine-tuned version of facebook/wav2vec2-base for Automatic Speech Recognition (ASR) in Puno Quechua (qxp).
Base pretrained model:
facebook/wav2vec2-base
Developers
This model was developed by:
- Johanna Cordova
- Elwin Huaman
- Adrian Gamarra Lafuente
Research Laboratories and Institutions
The development and training of this model involved the following research laboratories and institutions:
- ERTIM, Institut National des Langues et Civilisations Orientales (France)
- University of Cambridge (United Kingdom)
- Stanford University (USA)
Fine-Tuning Dataset
The model was fine-tuned using the following dataset:
- Dataset name: Mozilla Data Collective - Scripted Speech and Spontaneous Speech corpora
- Language: Puno Quechua
- Total duration: 65h
- Speech type: read speech / elicited speech
- Source: Mozilla Data Collective
Hardware and Training Configuration
The fine-tuning process was performed using the following hardware and software configuration:
- GPU: L40S, 48Gb
- Downloads last month
- 4
Model tree for QuechuaBase/wav2vec2-base-qxp-finetuned
Base model
facebook/wav2vec2-base