masuidrive/cv-corpus-17.0-ja-client_id-grouped

---
language:
- ja
license: cc0-1.0

tags:
- audio
- speaker diarization

source_datasets:
- commonvoice

task_categories:
- automatic-speech-recognition

annotations_creators:
- crowdsourced

language_creators:
- crowdsourced
---
# cv-corpus-17.0-ja-client_id-grouped

This dataset is a subset of the Common Voice dataset, filtered and grouped based on the client ID (treated as speaker ID).

## Dataset Details

- The dataset is derived from the Common Voice dataset.
- The original dataset is available at [Common Voice Dataset](https://commonvoice.mozilla.org/en/datasets).
- The dataset is grouped by client ID, which is treated as the speaker ID for this dataset.
- Each group is filtered to include only client IDs with a minimum of 30 samples and a maximum of 300 samples.
- The dataset is split into train and validation sets for each client ID group, with a ratio of 8:2.
- The same client IDs exist in both the train and validation sets.
- The dataset is split into batches of 1,000 samples and saved as Parquet files.

## Dataset Statistics

- Filtered client_id count: 649
- Filtered total entry count: 45,668
- Original total entry count: 89,310

## Sample Duration Distribution


The histogram shows the distribution of sample durations in the dataset.

## License

The Common Voice dataset is licensed under the Creative Commons Zero (CC0) license.

原始信息汇总

数据集概述

数据集名称

cv-corpus-17.0-ja-client_id-grouped

数据来源

• 该数据集是Common Voice数据集的一个子集。
• 原始数据集链接：Common Voice Dataset

数据处理

• 数据根据客户端ID（视为说话人ID）进行过滤和分组。
• 每个分组中，客户端ID的样本数量限制在30至300之间。
• 数据集按客户端ID分组，分为训练集和验证集，比例为8:2。
• 训练集和验证集中的客户端ID相同。
• 数据以1,000个样本为一批次，保存为Parquet文件格式。

数据集统计

• 过滤后的客户端ID数量：649
• 过滤后的总样本数：45,668
• 原始总样本数：89,310

样本时长分布

• 数据集中样本时长的分布情况可通过提供的直方图查看。

许可证

• 数据集遵循Creative Commons Zero (CC0)许可证。