gttsehu/Albayzin-2024-BBS-S2T-eval

Albayzin 2024 Bilingual Basque-Spanish Speech to Text (BBS-S2T) Challenge - Evaluation dataset

see Albayzin_2024_BBS-S2T_EvalPlan for a description of the challenge.

This is the evaluation data for the challenge.

The database consists of a single split:

1. eval : 12498 audio segments

How to download this database

1 - If you can handle yourself comfortably with Huggingface Datasets:

from datasets import load_dataset
ds = load_dataset("gttsehu/Albayzin-2024-BBS-S2T-eval")

The Dataset contains a single split:

DatasetDict({
    eval: Dataset({
        features: ['path', 'audio', 'sentence', 'speaker_id', 'language', 'PRR', 'length'],
        num_rows: 12498
    })
})

NOTE: speaker_id and PRR fields are noninformative in the actual version of the dataset.

2 - Manual download:

git clone https://huggingface.co/datasets/gttsehu/Albayzin-2024-BBS-S2T-eval

NOTE: git-lfs must be installed to be able to handle the download of the large tar files (which include the audio files).

Downloaded database structure:

Albayzin-2024-BBS-S2T-eval/
├── Albayzin-2024-BBS-S2T-eval.py
├── audio
│   └── eval_0.tar
├── languages.py
├── metadata
│   └── eval.tsv
├── README.md
└── release_stats.py

Untar the audio files:

ls Albayzin-2024-BBS-S2T-eval/audio/*.tar | xargs -i tar -xC Albayzin-2024-BBS-S2T-eval/audio -f {}

The metadata directory contains the index file for the evaluation split. The file contains five tab separated fields (NOTE: speaker_idand PRRfields are noninformative in the actual version of the dataset.):

1. The audio file path
2. The language of the segment (es: spanish, eu: basque and bi: bilingual)
3. The speaker id
4. The PhoneRecognitionRate indicating the quality of the transcription
5. The length of the segment (in seconds)
6. The transcription

path    language        speaker_id      PRR     length  sentence
eval/000a3dd1.mp3       es      0       100.00  8.89    yo le pregunto por qué no ocurre esto en la comunidad autónoma vasca por qué no ocurre esto espero que nos lo responda
eval/00158458.mp3       eu      0       100.00  3.82    eh bildu taldeak osakidetzako plantilla estrukturala
eval/00180e6f.mp3       eu      0       100.00  4.03    alderdiari bakarrik interesatzen zaizkienak euskal hiritar
eval/001cdd58.mp3       es      0       100.00  9.25    volver a actualizar ese esa investigación y la realización de ese diagnóstico yo estoy de acuerdo que
eval/00213e67.mp3       eu      0       100.00  5.47    gabiltza ez dago estrategiarik ez dago pertsonal finkorik ez dago
...