Work

Friday Night Lights Out: A Corpus of Radio Transmissions Broadcast by the Baltimore Police Department

Year: 2026
Type: preprint
Abstract: Friday Night Lights Out (FNLO) is a corpus of two-way radio transmissions broadcast by the Baltimore Police Department (BPD). The corpus is a subset of a larger archive of more than 91 million radio transmissions recorded on sixteen radio receivers between 2016 and 2022. Selected from that archive, the FNLO corpus targets a single, meaningful window: the second hour after sunset on each Friday in 2017.A team of undergraduate research assistants first transcribed a selection of radio transmissions. In a second phase of work, the research assistants then cut short audio segments keyed to important words and codes spoken in the transmissions, producing a set of ground-truth examples. The result of the effort is three linked components: 138,067 short audio recordings spanning the year, 25,893 human transcriptions of a subset of the transmissions, and 14,267 keyword clips (more than half of which correspond to radio transmissions in the set of 25,893 first-stage transcriptions). This document provides an explanation of recording techniques for the full archive of radio transmissions, the rationale for selecting transmissions for transcription, the transcription process itself (including the guidelines given to transcribers), the codes and conventions needed to understand the transcripts, and the 2017 policing context in which the transmissions were broadcast (a year framed by a federal consent decree, an emergent corruption scandal, and elevated violent crime).Because the transmissions contain sensitive information about possibly identifiable individuals, the authors hope to arrange in the future for the corpus to be made available to other researchers under appropriate safeguards. For now, it is available only to researchers at the first author’s institution by joining the research team. This document represents a step toward providing broader access to the corpus.
Language: English
Open Access status: green
OpenAlex ID: W7167486181
Cites:
Cited by:
Related to: 10