Speech-to-Text
Working Group
The Speech-to-Text Working Group focuses on automated and human-in-the-loop transcription of spoken language. The group shares practical approaches and discusses best practices for the use of Automatic Speech Recognition (ASR) workflows in Libraries, Archives, and Museums.
Building Collaboration and Knowledge Sharing Around Automatic Speech Recognition (ASR) in the LAM Community
Libraries, archives, and museums (LAMs) hold significant collections of recorded speech within their audio-visual materials. The automation of speech-to-text transcription represents an important development, enabling corpus creation for text mining, new forms of search and discovery, multilingual translation, enhanced accessibility, and more. With the growing availability of automatic speech recognition (ASR) systems, many LAM institutions are actively exploring and implementing speech-to-text workflows to caption and/or transcribe their audio-visual collections.
This group aims to share needs, experiences, and approaches; facilitate the dissemination of practical workflows and production pipelines; foster collaboration; and create opportunities for sharing data, models, and software in this area. The group also monitors the rapidly evolving ASR landscape and explores its implications for the LAM community.
We envision ASR as a topic of sustained and growing interest within the LAM sector. Accordingly, the group is intended to function as an ongoing interest group, open to participants of all experience levels, and capable of producing deliverables and collaborative outputs as needs and opportunities emerge over time.
Topics:
- ASR model performance across media genres and languages
- Tools, techniques, and workflows
- ASR data and training
- ASR in the AI and LAM landscapes
Monthly Call – The meetings takes place on the 4th Tuesday of each month with a duration of one hour.
16:00 UTC
The working group develops shared resources and accumulates a knowledge base to assist the LAM community in decision-making, approaches, and techniques for ASR.
- A repository of presentations on ASR
- A guide to ASR decision-making and best practices
- A metadata schema for recording the provenance of transcripts
- Methodologies and data for ASR evaluation
