Notes from a webinar on machine learning experiments in GLAMR.
The three talks covered recent AI initiatives at key cultural institutions in Australia and New Zealand. Common themes included:
The session was organised by AI4LAM chapter in Australia and New Zealand and co-hosted by Alexis Tindall and Ingrid Mason.
Adam Moriarty heads up a relatively new team with the aim to help people find connections with the collections. They have been working on a number of research projects that facilitated the bigger discussions about what is possible and what is ethical with machine learning.
Taking black and white photos from the collection and putting it through an algorithm to generate a colour version of the image.
Analysing a transcribed World War 1 Diary using IBM Watson (free tier) to track emotions throughout the text.
Is it better to have an AI created and captioned record online than no record at all?
Julia Hickie & Mark Raadgever
Magazines pose a different challenge to newspapers in that most are centrally published in Sydney and Melbourne. This means that the Marc place of publication can not be relied on as a trusted source of place. The National Library used machine learning to automate the process of assigning place metadata to magazines.
It’s worth noting that duplicate place names were not handled with automation.
Tatiana Antsoupova
A trial to map documents to retention and disposal schedule functions using text mining resulted in 66% accuracy however, machine learning algorithms resulted in 73-85% accuracy.
© 2022 Sally Baker