recordssaid

Recordssaid is a web-based archive and search platform for spoken language captured in audio recordings. It provides a centralized index of phrases with timestamps and speaker metadata, alongside machine-generated and human-curated transcripts. The project aims to support quote verification, linguistic research, and media accountability by enabling precise retrieval of spoken phrases.

Recordssaid originated in 2015 as a collaboration between linguists and digital archivists within the Global Speech Archives Initiative. The public beta launched in 2016, and an official release followed in 2017. The software is open source, with core components including the recordssaid engine and an API for programmatic access.

Content and metadata: The platform aggregates audio from public-domain sources and licensed collections, prioritizing materials with clear licensing. Each entry includes the spoken phrase text, the corresponding timestamp or time range, speaker information when identifiable, recording date and location, source collection, license, and a transcription quality flag.

Features: Users can perform full-text searches across transcripts, retrieve time-aligned segments for individual phrases, view speaker and context metadata, and contribute crowdsourced corrections. The platform supports export of data and integration via a REST API, and adopts interoperable metadata schemas.

Usage and impact: Researchers, journalists, educators, and archivists use recordssaid to verify quotes, study speech patterns, and track the dissemination of phrases across media. The project has influenced best practices for quote attribution and transparency in media archives.

Governance and challenges: As an open platform, it relies on community governance and clear licensing. Challenges include handling contemporary content, privacy concerns, and ensuring accurate attribution.