Devvrat024/Rural_Women_Bhojpuri

--- language: - bho - hi license: cc-by-sa-4.0 task_categories: - automatic-speech-recognition pretty_name: Rural Bhojpuri ASR Dataset dataset_info: features: - name: age_group dtype: string - name: audio dtype: audio: sampling_rate: 16000 - name: district dtype: string - name: duration dtype: float64 - name: job_type dtype: string - name: lang dtype: string - name: language dtype: string - name: prompt_text dtype: string - name: qualification dtype: string - name: scenario dtype: string - name: speaker_id dtype: string - name: state dtype: string - name: task_name dtype: string - name: text dtype: string splits: - name: benchmark num_bytes: 400115473 num_examples: 444 - name: train_real num_bytes: 460882675 num_examples: 400 - name: train_synthetic num_bytes: 39573449568 num_examples: 77967 download_size: 34075320626 dataset_size: 40434447716 configs: - config_name: default data_files: - split: benchmark path: data/benchmark-* - split: train_real path: data/train_real-* - split: train_synthetic path: data/train_synthetic-* --- # Rural Bhojpuri ASR Dataset ## Dataset Description This dataset is curated to foster the development of inclusive Automatic Speech Recognition (ASR) systems, with a special focus on the underrepresented voices of rural Bhojpuri women. It contains audio clips in both Bhojpuri and Hindi, collected from real-world and synthetic sources, designed to train and evaluate ASR models that can accurately recognize diverse speech patterns. This work is part of the research presented in the paper "Recognizing Every Voice: Towards Inclusive ASR for Rural Bhojpuri Women." ## How to Use The dataset can be easily loaded using the Hugging Face `datasets` library. ```python from datasets import load_dataset # Load the dataset dataset = load_dataset("ai4bharat/Rural__Women_Bhojpuri") # Access a specific split train_real_split = dataset["train_real"] # Print the first example print(train_real_split[0]) # The audio will be automatically decoded and resampled to 16kHz # Example: {'audio': {'path': '...', 'array': array([-0.00024414, -0.00048828, ...], dtype=float32), 'sampling_rate': 16000}, 'text': '...', ...} ``` ## Citation If you use this dataset in your research, please cite the following paper: ``` @misc{joshi2025recognizingvoiceinclusiveasr, title={Recognizing Every Voice: Towards Inclusive ASR for Rural Bhojpuri Women}, author={Sakshi Joshi and Eldho Ittan George and Tahir Javed and Kaushal Bhogale and Nikhil Narasimhan and Mitesh M. Khapra}, year={2025}, eprint={2506.09653}, archivePrefix={arXiv}, primaryClass={eess.AS}, url={[https://arxiv.org/abs/2506.09653](https://arxiv.org/abs/2506.09653)}, } ```language: - 博杰普尔语（Bhojpuri） - 印地语（Hindi） 许可证: CC BY-SA 4.0（知识共享署名-相同方式共享4.0协议） 任务类别: - 自动语音识别（Automatic Speech Recognition, ASR） 规范名称: 乡村博杰普尔语自动语音识别数据集 数据集信息: 特征: - 名称: age_group 数据类型: string - 名称: audio 数据类型: audio: 采样率: 16000 - 名称: district 数据类型: string - 名称: duration 数据类型: float64 - 名称: job_type 数据类型: string - 名称: lang 数据类型: string - 名称: language 数据类型: string - 名称: prompt_text 数据类型: string - 名称: qualification 数据类型: string - 名称: scenario 数据类型: string - 名称: speaker_id 数据类型: string - 名称: state 数据类型: string - 名称: task_name 数据类型: string - 名称: text 数据类型: string 拆分: - 名称: 基准集 字节数: 400115473 样本数: 444 - 名称: 真实训练集 字节数: 460882675 样本数: 400 - 名称: 合成训练集 字节数: 39573449568 样本数: 77967 下载大小: 34075320626 数据集总大小: 40434447716 配置项: - 配置名称: default 数据文件: - 拆分: benchmark 路径: data/benchmark-* - 拆分: train_real 路径: data/train_real-* - 拆分: train_synthetic 路径: data/train_synthetic-* # 乡村博杰普尔语自动语音识别数据集 ## 数据集说明 本数据集旨在推动包容性自动语音识别系统的研发，特别关注未被充分代表的乡村博杰普尔语女性群体的语音数据。数据集包含博杰普尔语与印地语的音频片段，采集自真实场景与合成数据源，用于训练和评估能够准确识别多样化语音模式的ASR模型。 本工作为论文《Recognizing Every Voice: Towards Inclusive ASR for Rural Bhojpuri Women》（《听见每一声：面向乡村博杰普尔语女性的包容性自动语音识别》）所呈现研究的一部分。 ## 使用方法 可通过Hugging Face的`datasets`库便捷加载本数据集。 python from datasets import load_dataset # 加载数据集 dataset = load_dataset("ai4bharat/Rural__Women_Bhojpuri") # 访问指定拆分集 train_real_split = dataset["train_real"] # 打印第一条样本 print(train_real_split[0]) # 音频将自动解码并重采样至16kHz # 示例输出：{'audio': {'path': '...', 'array': array([-0.00024414, -0.00048828, ...], dtype=float32), 'sampling_rate': 16000}, 'text': '...', ...} ## 引用说明 若您在研究中使用本数据集，请引用如下论文： @misc{joshi2025recognizingvoiceinclusiveasr, title={Recognizing Every Voice: Towards Inclusive ASR for Rural Bhojpuri Women}, author={Sakshi Joshi and Eldho Ittan George and Tahir Javed and Kaushal Bhogale and Nikhil Narasimhan and Mitesh M. Khapra}, year={2025}, eprint={2506.09653}, archivePrefix={arXiv}, primaryClass={eess.AS}, url={https://arxiv.org/abs/2506.09653}, }