IASA's 55th Annual Conference

➺ Using AI to make spoken-word content more accessible and discoverable
--
Mark Boas (Workshop)
--
Generative AI coupled with more traditional AI such as Automatic Speech Recognition (ASR) present unique opportunities for the categorisation, discoverability, accessibility and dissemination of spoken word media. In this workshop Mark Boas, CTO of TheirStory and co-founder of the Hyperaudio Project will discuss and demonstrate methods of exposing media in a form that facilitates search, sharing and even remixing using the open source tools he has been involved in developing for the last decade. The last 12 months has seen a sea change in the emergence of generative AI such as LLMs (Large Language Models), at the same time we’ve seen speech-to-text algorithms improve in accuracy and speed, while costs continue to fall.

His work at TheirStory, focussing on stories from underrepresented and marginalised communities, seeks to combine word-timed machine generated interactive transcripts with emerging generative AI to create metadata that facilitates the surfacing of content and its re-use in formats by students and other communities .

Believing that these technologies should be able to be leveraged by the wider community, much of this work adheres to open standards and has been released under open source licenses in forms that can be mixed, matched and integrated into other platforms.

Mark will take us on a whistle-stop tour of these new innovative technologies, explaining the concepts behind them – the pros and cons of each and inviting participants to try them out. Touching on the ethics of AI he’ll talk about the wider aim of engaging the oral-history archival community and fostering collaboration around open technologies to help expose this important aspect of our history.