I had a lovely week’s holiday last week and returned to work for one week only before I head off for a further two weeks. I spent most of my time this week working on the Speak For Yersel project implementing a huge array of changes that the team wanted to make following the periods of testing in schools a couple of weeks ago. There were also some new sections of the resource to work on as well.
By Tuesday I had completed the restructuring of the site as detailed in the ‘Roadmap’ document, meaning the survey and quizzes have been separated, as have the ‘activities’ and ‘explore maps’. This has required quite a lot of restructuring of the code, but I think all is working as it should. I also updated the homepage text. One thing I wasn’t sure about is what should happen when the user reaches the end of the survey. Previously this led into the quiz, but for now I’ve created a page that provides links to the quiz, the ‘more activities’ and the ‘explore maps’ options for the survey in question.
The quizzes should work as they did before, but they now have their own progress bar. Currently at the end of the quiz the only link offered is to explore the maps, but we should perhaps change this. The ‘more activities’ work slightly differently to how these were laid out in the roadmap. Previously a user selected an activity then it loaded an index page with links to the activities and the maps. As the maps are now separated this index page was pretty pointless, so instead when you select an activity it launches straight into it. The only one that still has an index page is the ‘Clever’ one as this has multiple options. However, thinking about this activity: it’s really just an ‘explore’ like the ‘explore maps’ rather than an actual interactive activity per se, so we should perhaps move this to the ‘explore’ page.
I also made all of the changes to the ‘sounds about right’ survey including replacing sound files and adding / removing questions. I ended up adding a new ‘question order’ field to the database and questions are now ordered using this, as previously the order was just set by the auto-incrementing database ID which meant inserting a new question to appear midway through the survey was very tricky. Hopefully this change of ordering hasn’t had any knock-on effects elsewhere.
I then made all of the changes to two other activities: the ‘lexical’ one and the ‘grammatical’ one. These included quite a lot of tweaks to questions, question options, question orders and the number of answers that could be selected for questions. With all of this in place I moved onto the ‘Where do you think this speaker is from’ sections. The ‘survey’ now only consists of the click map and when you press the ‘Check Answers’ button some text appears under the buttons with links through to where the user can go next.
For the ‘more activities’ section the main click activity is now located here. It took quite a while to get this to work, as moving sections introduced some conflicts in the code that were a bit tricky to identify. I replaced the explanatory text and I also added in the limit to the number of presses. I’ve added a section to the right of the buttons that displays the number of presses the user has left. Once there are no presses left the ‘Press’ button gets disabled. I still think people are going to reach the 5 click limit too soon and will get annoyed when they realise they can’t add further clicks and they can’t reset the exercise to give it another go. After you’ve listened to the four speakers a page is displayed saying you’ve completed the activity and giving links to other parts. Below is a screenshot of the new ‘click’ activity with the limit in place (and also the new site menu):
The ’Quiz’ has taken quite some time to implement but is now fully operational. I had to do a lot of work behind the scenes to get the percentages figured out and to get the quiz to automatically work out which answer should be the correct one, but it all works now. The map displays the ‘Play’ icons as I figured people would want to be able to hear the clips as well as just see the percentages. Beside each clip icon the percentage of respondents who correctly identified the location of the speaker is displayed. The markers are placed at the ‘correct’ points on the map, as shown when you view the correct locations in the survey activities. Question 1 asks you to identify the most recognised, question 2 the least recognised. Quiz answers are logged in the database so we’ll be able to track answers. Here’s a screenshot of the quiz:
I also added the percentage map to the ‘explore maps’ page too, and I gave people the option of focussing on the answers submitted from specific regions. An ‘All regions’ map displays the same data as the quiz map, but then the user can choose (for example) Glasgow and view the percentages of correctly identified speakers that respondents from the Glasgow area identified, thus allowing them to compare how people in each area managed to identify speakers in the areas. I decided to add a count of the number of people that have responded too.
The ‘explore maps’ for ‘guess the region’ has a familiar layout – buttons on the left that when pressed on load a map on the right. The buttons correspond to the region of people who completed the ‘guess the region’ survey. The first option shows the answers of all respondents from all regions. This is exactly the same as the map in the quiz, except I’ve also displayed the number of respondents above the map. Two things to be aware of:
Firstly, a respondent can complete the quiz as many times as they want, so each respondent may have multiple datasets. Secondly, the click map (both quiz and ‘explore maps’) currently includes people from outside of Scotland as well as people who selected an area when registering. There are currently 18 respondents and 3 of these are outside of Scotland.
When you click on a specific region button in the left-hand column the results of respondents from that specific region only are displayed on the map. The number of respondents is also listed above the map. Most of the regions currently have no respondents, meaning an empty map is displayed and a note above the map explains why. Ayrshire has one respondent. Glasgow has two. Note that the reason there are such varied percentages in Glasgow from just two respondents (rather than just 100%, 50% and 0%) is because one or more of the respondents has completed the quiz more than once. Lothian has two respondents. North East has 10. Here’s how the maps look:
On Friday I began to work on the ‘click transcription’ visualisations, which will display how many times speakers have clicked in each of the sections of the transcriptions they listen to in the ‘click’ activity. I only managed to get as far as writing the queries and scripts to generate the data, rather than any actual visualisation of the data. When looking at the aggregated data for the four speakers I discovered that the distribution of clicks across sections was a bit more uniform that I thought it might be. We might need to consider how we’re going to work out the thresholds for different sizes. I was going to base it purely on the number of clicks, but I realised that this would not work as the more responses we get the more clicks there will be. Instead I decided to use percentages of the total number of clicks for a speaker. E.g. for speaker 4 there are currently a total of 65 clicks so the percentages for each section would be:
|11%||Have you seen the TikTok vids with the illusions?|
|9%||I just watched the glass one.|
|17%||The guy’s got this big glass full of water in his hands.|
|8%||He then puts it down,|
|8%||takes out one of those big knives|
|6%||and slices right through it.|
|6%||I sometimes get so fed up with Tiktok|
|8%||– really does my head in –|
|8%||but I’m not joking,|
|14%||I want to see more and more of this guy.|
(which adds up to 101% with rounding). But what should the thresholds be? E.g. 0-6% = regular, 7-10% = bigger, 11-15% even bigger, 16%+ biggest? I’ll need input from the team about this. I’m not a statistician but there may be better approaches, such as using standard deviation and such things.
I still have quite a lot of work to do for the project, namely: Completing the ‘where do you think the speaker is from’ as detailed above; implementing the ‘she sounds really clever’ updates; adding in filter options to the map (age ranges and education levels); investigating dynamically working out the correct answers to map-based quizzes.
In addition to my Speak For Yersel work I participated in an interview with the AHRC about the role of technicians in research projects. I’d participated in a focus group a few weeks ago and this was a one-on-one follow-up video call to discuss in greater detail some of the points I’d raised in the focus group. It was a good opportunity to discuss y role and some of the issues I’ve encountered over the years.
I also installed some new themes for the OHOS project website and fixed an issue with the Anglo-Norman Dictionary website, as the editor had noticed that cognate references were not always working. After some investigation I realised that this was happening when the references for a cognate dictionary included empty tags as well as completed tags. I had to significantly change how this section of the entry is generated in the XSLT from the XML, which took some time to implement and test. All seems to be working, though.
I also did some work for the Books and Borrowing project. Whilst I’d been on holiday I’d been sent page images for a further ten library registers and I needed to process these. This can be something of a time-consuming process as each set of images needs to be processed in a different way, such as renaming images, removing unnecessary images at the start and end, uploading the images to the server, generating the page images for each register and then bringing the automatically generated page numbers into line with any handwritten page numbers on the images, which may not always be sequentially numbered. I processed two registers for the Advocates library from the NLS and three registers from Aberdeen library. I looked into processing the images for a register from the High School of Edinburgh, but I had some questions about the images and didn’t hear back from the researcher before the end of the week, so I needed to leave these. The remaining registers were from St Andrews and I had further questions about these, as the images are double-page spreads but existing page records in the CMS treat each page separately. As the researcher dealing with St Andrews was on holiday I’ll need to wait until I’m back to deal with these too.
Also this week I completed the two mandatory Moodle courses about computer security and GDPR, which took a bit longer that I thought they might.