Bulgarian audio dataset for speech recognition 20 hours (2/4)

Specifications: - Each user has a unique ID across the entire dataset. - Maximum four hours of speech per person in the dataset. - Speech is recorded and transcribed on separate tracks. - High-quality transcriptions come with the data in JSON format. - No noise and high-quality recordings with both male and female speakers. - Metadata includes: gender, age, and location. - License terms: you pay once and you can use the data commercially in your products, but you cannot resell the data.

搜集汇总

背景与挑战

背景概述

该数据集为保加利亚语语音识别提供20小时高质量录音，包含男女说话者的无噪声语音及JSON格式转录，每人最多录制四小时。数据附带说话者性别、年龄和位置等元信息，并允许一次性付费后在产品中商业使用，但禁止转售。