SEP-28k
收藏资源简介:
SEP-28k数据集是由纽伦堡工业大学创建的一个大型语音数据集,包含约28000个3秒长的语音片段,这些片段来自围绕口吃话题的385个播客节目,并标注了五种不同的口吃事件类型。数据集的创建过程涉及从播客中提取片段并进行手动标注。该数据集主要用于口吃检测系统的研究和开发,旨在解决口吃检测中的数据稀缺问题,并提高语音识别系统的包容性。
The SEP-28k dataset is a large-scale speech dataset created by Technische Universität Nürnberg. It contains approximately 28,000 3-second-long speech segments derived from 385 podcast episodes centered around the topic of stuttering, with annotations for five distinct types of stuttering events. The dataset creation process involves extracting segments from podcasts and performing manual annotation. This dataset is primarily used for the research and development of stuttering detection systems, aiming to address the data scarcity issue in stuttering detection and improve the inclusivity of speech recognition systems.

- 1SEP-28k: A Dataset for Stuttering Event Detection From Podcasts With People Who Stutter苹果 · 2021年



