Monday was a holiday this week and on Tuesday I worked on the Burn Supper map.  I updated the narrative summary so that it features percentages based on the number of suppers that supplied the data type rather than the overall total.  This means that (for example) Whisky now has the percentage 88% rather than 57%.  The new summary layout is shown below:

I then ran an import of new suppers, taking the total of suppers up to 1044, and as the facts and figures are all dynamically generated these updated to reflect the new data. I had to do some manual tweaking of the data, for example some countries had typos and needed fixing, plus I needed to process and upload all new images that were associated with the new suppers.  Other than future imports of new data I think the Burns Supper map is now pretty much complete.

On Wednesday I focussed on the Playbills project.  Last week I’d executed my script to process performers, and this left a few hundred that needed to be manually checked.  I started to work on these myself initially, but it was pretty slow going, especially when checking data by querying the database directly.  Instead I decided to create a new spreadsheet that includes all of the fields needed for checking (e.g. the role information as well as the performer details) and it should be quicker to edit this.  I sent it, along with detailed instructions on how to update the data, to Deven and hopefully she’ll be able to work through the list in an hour or two.

I met with Deven to discuss our next steps for the project during the afternoon, and before this meeting I prepared a list of discussion points.  As we would ideally like to know the genders of all roles in the data (so it will be possible, for example, to ascertain when female performers take on male roles) but as roles are often just names without titles assigning gender would require manual checking.  Instead I wondered whether we could get AI to assign gender and then leave us with only the trickier ones that are less clear.

As a quick experiment I passed a short list of roles from one play to ChatGPT and asked it to ascertain the gender of each.  The prompt I gave was “for the following list of names work out whether each is male, female or unknown: Edmund (the Blind Boy), Stanislaus, Oberto, Rodolph, Kalig, Molino, Starrow, Elvina, Lida” and the response was: “Edmund (the Blind Boy) — Male,  Stanislaus — Male,  Oberto — Male, Rodolph — Male, Kalig — Unknown, Molino — Unknown, Starrow — Unknown, Elvina — Female, Lida — Female”

So two thirds of roles were correctly assigned a gender and there were no mistakes, leaving one third that would need manual checking, which I think is looking fairly promising.

On Thursday I continued to develop the interactive map for the HiMuJe Malabar project.  I sorted out the type / subtype categorisation in the legend to include both types and subtypes, as shown in the following screenshot:

Subtypes are indented within the type now and any places with a type but no subtype (e.g. Sri Lanka) are now appearing.  Currently all types and subtypes appear in the legend, even if they have no associated places, mainly so we can see what the full list will look like.  It is rather long and I may need to add in a scrollbar, although we may also want to rework the categorisation too – e.g. we have ‘Town’ as both a type and a subtype of ‘Settlement’, plus we have two occurrences of both ‘Hinterland’ and ‘Backwater’.

I was intending the counts beside the types to be a total of all places categorised by the respective subtypes, but some places only have a type and no subtype so the counts represent these instead (e.g. the ‘4’ beside ‘Region’ shows the number of places that have ‘Region’ and no subtype).  There are also 5 places that have a subtype within ‘Region’ in addition to this but I’m not sure how best to represent this without confusing people.  We could have something like ‘Region (4+5)’ or ‘Region (9)’ but both of these seem a bit unclear to me.  I was also thinking of having the checkbox beside each main type select / deselect all subtypes, but if we did this it wouldn’t be possible to just display the main type without its subtypes.  These issues need further consideration.

I also implemented the record pop-up that now appears when you press on a marker or polygon, as you can see in the following screenshot:

The popup header displays the ‘preferred name’ for the place and the ‘general information’ tab features the ID, all names and their languages, the category and subcategory (I guess I should standardise this to ‘type’ and ‘subtype’ to avoid confusion) and any supplied description. The ‘references’ tab shows a count of the number of references to the placename in the source texts and the content of the tab lists the filenames and snippets for each reference, with the actual text highlighted in yellow, as shown in the following screenshot:

We should probably have actual titles for the source texts rather than filenames, but these are not included in the output and is maybe something to add in, along with references to specific lines / pages.  Another possible issue is that I am aware some of the text will be read right to left and at the moment all text is just displayed as ‘prefix+extract+suffix’.  We might need a flag in the data for when the text should instead be ‘suffix+extract+prefix’ (or I guess set the direction to ‘rtl’ in the stylesheet).  When we have any images I’ll add these as a further tab, but we don’t have any yet.  Also, I haven’t implemented the ‘Share / Cite’ tab yet.

The final thing I’ve implemented is an alternative categorisation for the map, based on the source texts the place is referenced in.  In the data I was working with there are only four places that have references, and I’ve added a further ‘No source’ category that all other places are added to, as the following screenshot demonstrates:

I renamed the ‘Place’ menu in the left-hand menu to ‘Place Categorisation’ and there is now an option to switch the categorisation from type to source text.  I’ll add in ‘language’ and ‘frequency of reference’ next week, all being well.  After I sent an update to the team, Christian, the technical person for the digital edition, informed me that a new output of the data was available that featured many more references and other updates.  I therefore replaced the data in the map with the new version and it made a huge difference – the list of cats and subcats is now shorter, there are many more source texts (although this demonstrated that I have some further work to do with the legend for source text categorisation) and many more places with references.  I’ll continue with this next week.

During the week Katie Halsey, the PI of the Books and Borrowing project, contacted me to ask for some help in creating some queries of the data for the monograph that she and Matt are writing.  These involved working out which books and authors were found at ten or more libraries, and which books and authors were borrowed in every decade from the 1750s to the 1830s.  I looked into this on Friday, and it took most of the day to work on it, partially because it’s been a while since I worked with the data and it took some time to remember how everything fitted together.  However, I managed to produce the data required data and I sent it to Katie and Matt in four spreadsheets.

Also on Friday I did some work for the Dictionaries of the Scots Language, setting up a new user account that will be used by some interns that are starting with the project over the summer, sorting out access to the Google Search Console and removing the user survey.