It was another four-day week due to Easter Monday, and I spent a fair amount of my time working for Speak For Yersel.  For our abstract for DH2025, I received some new, high-resolution images of the graphs from co-author Marc Barnard and after a little more tweaking and dealing with the submission process I was able to complete the submission process.  I still need to actually sign up for the conference, though, which is something I’m hoping to be able to do next week.

One of the reviewers of our abstract had made a comment about user fatigue and enquired about how many users actually completed the surveys and this prompted me to undertake some investigation.  I created a series of database queries that for each user extracted counts of the number of submitted answers in the morphology, lexis and phonology surveys.  I then exported the data into Excel spreadsheets, one per region (Scotland, Northern Ireland, Ireland and Wales) and sent this data on to Jennifer in case it was of some use.

However, I did have to point out that the counts for each user are of answers submitted and some questions allow for multiple answers to be selected so there’s not an exact 1:1 relationship between the figures and the number of questions.  Also, a user can complete a questionnaire more than once (or quit mid-way and then begin a second time).   I noticed that there were a few users who have submitted many more answers than there are questions.  For example, in the NI data there is a user who has submitted 102 morphology, 105 lexis and 76 phonology answers, even though there are only 34, 35 and 29 questions respectively.  It’s not possible to ascertain exactly what prompted the user to submit so many answers, but it is perhaps a case of someone passing an iPad round a group of people.

I then decided to experiment with visualising the data, and for this I used the Highcharts library (https://www.highcharts.com/).  There are lots of interesting visualisation that can be made with the data, and my first attempt was to generate a stacked column chart (https://www.highcharts.com/demo/highcharts/column-stacked).  I decided that it would be useful to visualise all of the answers submitted by each user over time so wanted to plot each user as a column, with columns on the x-axis arranged chronologically by date of user account creation.  Each column would then be split into three coloured sections showing the number of answers submitted across the three surveys (morphology, lexis and phonology).

I wrote a little script to generate the JSON data that Highcharts can easily work with, and adapted the Highcharts example linked to above to work with the new data on my local PC.  The resulting graphs include a lot of columns (6379 for Scotland, 785 for Northern Ireland, 350 for Ireland and 1487 for Wales) and everything does get rather squashed together and some gaps can get lost, but what is interesting is how these graphs show the overall pattern of submissions.  On the whole most users across all regions made a decent stab at completing all surveys, as there are clear bands of colours, admittedly with some variation and gaps.  The graphs also show the outliers, especially the people who submitted many more answers than there are questions.  One user in Scotland in particular seems to have gone a bit crazy, and this then causes the rest of the graph to be rather squashed.

Here’s the graph for Northern Ireland, showing a few users who submitted many more answers than there are questions, a few users who submitted very little, and an overall pattern showing users making a decent stab at completing all three surveys.

The pattern for Ireland is broadly similar, and with less users it’s easier to view the individual columns:

Wales has considerably more users, which means individual columns can get lost, but it’s still possible to get an overall sense of user completion rates:

For Scotland we have an awful lot more users, plus as mentioned earlier one user who submitted huge numbers of answers, which results in a graph that squashes up all of the other data:

I had to regenerate the Scotland data and graph as after my first attempt I realised that the Scotland resource features more than just the surveys but also features follow-on questions, quizzes and other activities and I hadn’t filtered all of this out.  I was also interested to expand the x-axis to allow a more nuanced view of the data, and also to place a maximum extent on the y-axis to avoid one user affecting the display of data for all other users:

The resulting graph above (which you’ll need to open to view properly) demonstrates a lot more variation in user submissions that are lost from the smaller graph and demonstrates how the smaller graph with its big blocks of solid submissions doesn’t reflect reality.  Having said that, the overall picture still shows a decent number of users submitting a complete or near-complete set of answers.  Of course there is still much that could be done with the data.  Even using the same stacked column graph, rearranging the users by number of submissions would be interested, and may demonstrate how the overall picture when presented in date order could possibly obscure the number of users who didn’t submit much data.  But that’s for another time.

Also this week I made a few changes to the Speech STAR resource including adding a link to the ‘in clinic’ site form the top level tabs for Seeing Speech, Dynamic Dialects and the STAR site, as you can see here: https://www.seeingspeech.ac.uk/speechstar/.  I also had a chat with Eleanor about the visibility of the STAR sites in search results.

I also made a few further updates to the VARICS lookup test that’s still in development, including updating the ‘how to’ guides and linking to them from the longer textual descriptions of measurements.

On Friday I had a Teams meeting with Katie, Matt and Kitt from the Books and Borrowing project to discuss the new versioning system I’m going to develop.  We agreed that I will develop the full versioning system I specified in the document I sent around two weeks ago, and I’ll aim to get started on this in the next week or two.