LibriLong

Speech is a tool for communication that conveys private information from one agent to another. In addition to the legitimate message, speech contains plenty of bundled side information, such as state of health, emotions, identity, and background, most of which is private. Speech technology, which processes, transmits, or stores speech, can thus expose users to a multitude of threats, such as price gouging, stalking, and identity theft. Protecting user privacy in speech applications is morally important, but since threats to users also reduce user satisfaction, it is also important for business.LibriLong data set is a collection of LibriVox public audio books with the aim of having a long sequence of speech data that resembles a streamed speech data from one individual speaker. The data set is primarily used for privacy analysis for streamed speech data. LibriLong consists of 64 different audio books each from an individual speaker where it contains 31 male and 33 female speakers. The length of speech data per speaker ranges from 2h:29m to 4h:54m hours (the shortest to the longest).