Скачать или смотреть AI Metadata generation as shortcut to access

FIAT/IFTA World Conference 2023
Locarno, Switzerland
Presentation date: October 20th, 2023
Speakers: Karen Brøcker (The Royal Danish Library)

-- Abstract --

In our current practice, workflows that give access to radio and tv programs are either mostly manual or mostly automated. For the homogeneous collections, they tend to be automated, but for the heterogeneous collections, human judgement is needed. In this presentation, I explore the ways in which AI metadata generation can act as a bridge between the two scenarios. For instance, can we give access to programs with unknown titles by making them searchable using speech to text metadata generation? Can we avoid manual cutting of longer chunks of tape by adding speech to text metadata, thereby making the larger chunks searchable?

These solutions fall short of full automation of workflows, but assist the manual process and enables more material to be processed on a shorter timeframe than would otherwise be possible. The presentation reports on firsts attempts at using speech to text metadata generation on The Royal Danish Library’s radio/tv collections and present potential future use-cases.