Personal researcher notes

The week started with a great discussion with BT about the personal researcher outputs. Elspeth felt that the descriptions generated presented a good case for a volunteer-led project within the archive where they would get more people to take notes on uncatalogued files and automatically create descriptions. When I questioned why they would not rely on the volunteers to write the descriptions themselves, both Elspeth and James said they felt GPT had a more consistent tone of voice appropriate for archival descriptions.

I hope to send out the feedback forms to volunteers next week, the delay has been caused by some file upload issues with ChatGPT, which I hope are resolved by next week. For this, I’m somewhat in the hands of OpenAI.

Monday’s chat also resulted in my favourite quote of the project so far, from James: “For our work, broader economic trends are the threat … but this [digital technique] … it enables us to do other things.”

Not too shabby for a Monday at 11am!

Gender

Anna-Maria and I have reflected that the gender work may have gone as far as it is able to go for now. We will be bringing some questions to the first investigations meeting in March to ask for the wider groups’ opinion, but the current thinking is that we are able to write up the following:

• An evaluation of genderize as a tool
• A discussion of other gender-detector tools currently available, what they do (in theory), and what this looks like in practice when the tools themselves are very hard to gain access to
• What an ideal tool would look like, if we could create one: this would sit somewhere between all three tools we have evaluated (genderize, the gender bias decoder, and Lucy Havens’ annotation model), while also bringing in domain expertise
• A reflection on institutional nervousness about a computer detecting bias in cataloguing practices
• Suggestions for how we could begin uncovering other protected characteristics, especially class and race which are so key to industrial history

Comms and computer vision

It was great to see how Kaspar has been developing the Heritage Weaver work, and we are hoping to have an in-person meeting to use the tool in two weeks’ time. While the principal aim of this work is to see what information is most useful for collections linkage, an accidental consequence of Kaspar’s work with text search is that it may be digitally possible to trace histories of collecting and curatorial voice throughout collections.

Once the comms dataset has been ingested, I’m optimistic we will start to see some interesting results, and I think this kind of visual linkage would fit well into the final exhibit.

Postal Archive

Jon and I visited the postal archive on Wednesday, which was both fruitful and frustrating in equal measure – although perhaps that is every archival trip ever?

It was great to see that the ‘Post Office Guides for Bradford and District’ contain tabulated information about where Bradfordians could speak to, what means they could use to speak, and how long it would take to receive goods from abroad. It was also surprising to find detailed records of which trams had post boxes installed, as well as where post boxes and offices could be found throughout Bradford.

The frustration comes from the fact that we only have this information for 1902, 1903, 1905, 1906, 1907, and 1915. These Post Office Guides were sold at every Bradford Post Office, every year, for 1 shilling, and in an ideal world we would have a guide from every year. The archivist at the Postal Museum suspects any others held by them may have been lost in the second world war, and while there’s still hope others are held somewhere else, I’m not sure we will find them within this project.

The information from the archive could be incorporated into the circulars work, or used to map (geospatially or isochronally) Bradford’s international communications connections. Most of the guides are now available on Box, alongside a newspaper which reveals something of the connection between Bradfords’ textile industry and communication networks: Postal Museum Images | Powered by Box