Cnam-LMSSC/multilingual_librispeech_spanish_phoneme

---
annotations_creators:
- machine-generated
language:
- es
language_creators:
- crowdsourced
license: cc-by-4.0
multilinguality:
- monolingual
size_categories:
- 1K<n<10K
- 10K<n<100K
source_datasets:
- facebook/multilingual_librispeech
task_categories:
- automatic-speech-recognition
- text-to-speech
pretty_name: Multilingual LibriSpeech Spanish (Phonemized & Curated)
tags:
- audio
- speech
- phonemes
- ipa
- spanish
- mls
- librispeech
- cnam-lmssc
---

# Multilingual LibriSpeech Spanish Phoneme

## Dataset Description

- **Repository:** [Cnam-LMSSC/multilingual_librispeech_spanish_phoneme](https://huggingface.co/datasets/Cnam-LMSSC/multilingual_librispeech_spanish_phoneme)
- **Point of Contact:** [Cnam-LMSSC Team](https://huggingface.co/Cnam-LMSSC)

### Dataset Summary

This dataset is a **curated version of the Spanish subset of Multilingual LibriSpeech (MLS)**, enriched with a phonetic transcription column (`phoneme`).

The **Laboratoire de Mécanique des Structures et des Systèmes Couplés (Cnam-LMSSC)** created this version to facilitate research into Spanish acoustic modeling, phoneme recognition, and speech synthesis. It builds upon the high-quality audio derived from LibriVox audiobooks found in the original MLS dataset.

### Supported Tasks
- **Phoneme Recognition:** Training models to map audio features directly to IPA phonemes.
- **Automatic Speech Recognition (ASR):** Robust speech-to-text tasks using audiobook data.
- **Text-to-Speech (TTS):** Utilizing the high-quality, read-speech nature of LibriSpeech for synthesis.

### Languages
The dataset contains audio in **Spanish (`es`)**.

## Dataset Structure

### Data Instances

A typical data point contains the audio, metadata about the speaker and chapter (derived from LibriVox), the original text, and the generated phonetic transcription.

```python
{
'id': '1234_5678_00001',
'audio': {
'path': '1234_5678_00001.flac',
'array': array([-0.012, 0.045, ...], dtype=float32),
'sampling_rate': 16000
},
'text': "pestañeado quedóse en donde estaba apoyándose apenas en su muleta",
'phoneme': "pestaɲeaðo keðose en donde estaβa apoʝandose apenas en su muleta",
'speaker_id': 1234,
'chapter_id': 5678,
}

annotations_creators:
- 注释来源：机器生成
language:
- 西班牙语（`es`）
language_creators:
- 语言生成来源：众包
license: 知识共享署名4.0国际许可协议（CC BY 4.0）
multilinguality:
- 语言类型：单语
size_categories:
- 1K<n<10K
- 10K<n<100K
source_datasets:
- facebook/multilingual_librispeech
task_categories:
- 自动语音识别
- 文本到语音转换
pretty_name: 展示名称：多语言利布雷斯皮奇西班牙语版（音素化与精选版）
tags:
- 音频
- 语音
- 音素
- 国际音标（IPA）
- 西班牙语
- mls
- librispeech
- cnam-lmssc

# 多语言利布雷斯皮奇西班牙语音素数据集

## 数据集说明

- **仓库地址**：[Cnam-LMSSC/multilingual_librispeech_spanish_phoneme](https://huggingface.co/datasets/Cnam-LMSSC/multilingual_librispeech_spanish_phoneme)
- **联络人**：[Cnam-LMSSC团队](https://huggingface.co/Cnam-LMSSC)

### 数据集概览

本数据集是**多语言利布雷斯皮奇（Multilingual LibriSpeech，简称MLS）西班牙语子集的精选版本**，新增了音素转录列（`phoneme`）。

**法国国家工艺学院耦合系统与结构力学实验室（Cnam-LMSSC）**制作该版本的目的是推动西班牙语声学建模、音素识别与语音合成领域的研究。本数据集基于原始MLS数据集中来自LibriVox有声书的高质量音频构建。

### 适用任务

- **音素识别**：训练模型将音频特征直接映射至国际音标（IPA）音素。
- **自动语音识别（ASR）**：利用有声书数据完成鲁棒性语音转文本任务。
- **文本到语音转换（TTS）**：依托LibriSpeech高质量朗读语音的特性，实现语音合成。

### 语言说明

本数据集的音频语言为**西班牙语（`es`）**。

## 数据集结构

### 数据样本

典型数据样本包含音频、源自LibriVox的说话人与章节元数据、原始文本以及生成的音素转录结果。

python
{
'id': '1234_5678_00001',
'audio': {
'path': '1234_5678_00001.flac',
'array': array([-0.012, 0.045, ...], dtype=float32),
'sampling_rate': 16000
},
'text': "pestañeado quedóse en donde estaba apoyándose apenas en su muleta",
'phoneme': "pestaɲeaðo keðose en donde estaβa apoʝandose apenas en su muleta",
'speaker_id': 1234,
'chapter_id': 5678,
}