Spotify Podcast Dataset

Spotify Podcast Dataset是由Spotify创建的大规模音频数据集，包含约10万集播客，总时长超过4.7万小时，并附有完整的自动语音识别（ASR）转录文本。该数据集涵盖了从专业制作到业余创作的广泛内容，包括新闻、故事讲述、体育、健康等多个领域，以及不同的结构格式和发言者数量。数据集的创建过程中，采用了随机抽样和多重筛选标准，确保了数据的质量和多样性。该数据集主要用于信息检索和自然语言处理领域的研究，旨在解决音频内容分析和处理中的挑战。

Spotify Podcast Dataset is a large-scale audio dataset created by Spotify, containing approximately 100,000 podcast episodes with a total duration of over 47,000 hours, and is accompanied by complete automatic speech recognition (ASR) transcriptions. This dataset covers a broad spectrum of content ranging from professionally produced works to amateur creations, spanning multiple domains including news, storytelling, sports, health and more, as well as diverse structural formats and varying numbers of speakers. During its construction, random sampling and multiple screening criteria were adopted to ensure data quality and diversity. This dataset is primarily used for research in the fields of information retrieval and natural language processing, aiming to address challenges in audio content analysis and processing.

搜集汇总

背景与挑战

背景概述

Spotify Podcast Dataset是由Spotify提供的大规模音频数据集，包含约10万集播客，总时长超过4.7万小时，并附有自动语音识别（ASR）转录文本，内容涵盖新闻、故事讲述、体育、健康等多个领域，具有高度多样性。该数据集通过随机抽样和筛选确保质量，主要用于信息检索和自然语言处理研究，以解决音频内容分析和处理的挑战。