I added a new project worksheet to Machine Learning for Kids today.
It has step-by-step instructions for how to make a quiz show game in Scratch that uses a machine learning model to understand questions on any topic the student chooses, and find the answer in Wikipedia pages.
It’s a fun little project, super simple to make, and works surprisingly well. It doesn’t get every question right, but it does a lot better than I expected.
I don’t normally write blog posts when I write new ML for Kids worksheets, but this one was a bit interesting.
The topic I chose to demo the project was the Wimbledon tennis championships – asking questions such as what colour balls they use at Wimbledon, when the retractable roof was first introduced, the spectator capacity of Centre Court, and so on.
The reason I wanted to mention it is because that was a bit of a callback.
In June 2016, I shared a demo of something I was working on, called Watson Retrieve & Rank. In that demo, I showed how to train a machine learning system to find answers to questions in a corpus of documents. To do that I also used questions about the Wimbledon tennis championship, retrieving answers from a corpus of documents about Wimbledon.
These two demos give a nice snapshot into how far machine learning tech has progressed in the last five years.
The fact that you can now do this in JavaScript in a web browser without any dependencies on web services (the only Internet access in the project is where it retrieves the Wikipedia page – everything else is done on the student’s own computer running in the web browser) is just amazing.
(To be clear, these two demos are not doing exactly the same job – in Retrieve and Rank, the QA system was finding an answer in a large corpus of documents, whereas in this student Scratch project, the QA system is finding an answer in a single page. But it’s similar. And still super impressive to me.)
youtu.be/SveIksv7V9E (June 2016)
And that wasn’t even the first time I worked on a question answering system that could answer questions about Wimbledon. In 2015, I worked on a project for Wimbledon using something called Watson Engagement Advisor:
… Watson Engagement Advisor provides related insights and historical context. Wimbledon staff will be able to pose questions in natural language as if they had the world’s best tennis expert on-hand, and share these insights with fans via social media and the Wimbledon digital platforms … [This] will bring unprecedented analysis … For example, Wimbledon staff will be able to quickly surface information and insights about interesting or record-breaking player and match statistics
(I don’t have any screenshots of that project to hand, but you can see some in this blog post.)
Essentially, I’ve spent a lot of time making question answering systems about the Wimbledon championships. So it was a fun bit of question answering nostalgia for me to revisit that again, this time to enable children to do it!