Overview of the heritage weavings, TagLab, and visualising oral histories investigations before going into a conversation about thoughts and questions emerging from the cluster of work.

Domain experts

Sally from Leicester planning to use the tool to try and do some research and reflect on the use of the tool. Problem is that tools are, by definition, generalist. We need to understand how to apply them to specialist case studies. (Stef)

TagLab - methodology works well for social media, but it isn’t working well for catalogues. The domain expert’s role in this work is highlighting where the issue is - not that it’ll be an ongoing requirement - but more that it is needed to refine the tools/models for usage.

Kaspar - looking to work with Tasha to develop the human-in-the-loop aspect of the datasets and models to be more specific.

Daniel - domain expertise is very subjective. His work on TagLab is being approached very differently by Tasha - likely to be because of different data sets, but also because of different background experiences and general approaches.

Jane - relying on one or two domain experts is great for getting certain expertise, but it might also inadvertently build very specific models/perspectives. Need to think clearly about positionality.

Infrastructural issues

How can you build a large model without a GPU (Jack) - echoed by Stef - they are using a Uni of Leicester high performance computer with a 100 GPUs.

Kunika - some technical issues that used to be faced with being able to read newspapers are not really a problem anymore - it is now more a case of understanding and becoming used to the notions and nuances of the tools alongside a deeper understanding of the context from domain experts.

Daniel - sharing the insights of usable tools can be hugely productive - when someone understands how they can make use of the tool themselves (rather than technical presentations) then it becomes more accessible and we start to see potential impact.

Reflective documentation should be happening on positionality - AMS going to be proposing an investigation walking alongside a few investigations to think about positionality in our approaches. We also need to think about sustainability in our usage of computational power. (Alex has suggested Jo Walton(?) to do some sort of sustainability review etc).

Collaboration

Seeing links between each of the approaches - potential to look at how to enrich object descriptions through some of the approaches being done in the project.

What do we think about how the sector would respond to enhancing catalogue data?

Think more about this on 11th December at an automated collections linkage investigation meeting.