A team of cognitive scientists is trying to develop a ‘visual grammar’ of letter shapes. Would you like to help them? You can do so by playing a new online game developed by the research team. Players compete to develop rules that describe the shapes of letters in a wide range of writing systems.
Using machine learning tools, the research team will analyse the rules submitted by players. The team are looking to develop reliable and informative classification rules that will provide a reproducible and transparent typology for describing letter shapes across fields such as linguistics and visual psychology. The first question the team will try to answer is the following: how can letters in a script differ from one another enough to be distinct without being too difficult to process.
Playing the game
The game is on an online app called ‘glyph’ at https://glyph.shh.mpg.de. Here’s how you play.
- You select one of the available alphabets (or other scripts).
- Next, you create a rule describing some of the characters in that alphabet. The rule must describe at least 2 characters, but not more than a maximum. The maximum seems to be just less than half of the total number of characters in the set. You select all the characters complying that rule, and then submit a written description of the rule.
- After a few minutes, the app allows you to start playing that rule.
- When you play a rule, the game shows you the name of the rule and all of the characters in that alphabet. The characters are in a random order. If you succeed in selecting the same characters as when you submitted the rule, you pass the test and the app gives you points. If you do not succeed within 3 attempts, you fail the test and get no further attempts at playing that rule.
- To reward the most informative rules, the number of points equals the number of characters selected when you set up the rule. As an incentive for developing new rules, there are additional points for creating a rule not yet created by anyone else. None of my rules so far are unique, so I don’t know how that part of the scoring works!
- The game offers a selection of a small number of diverse writing systems. As you win points, the app makes more alphabets available.
My home page currently shows 15 scripts available. I have played 8 rules on Old Permic (scoring 21points) and 8 rules also on Zanzabar square (86 points). I haven’t created any rules for the other 13 languages available.
If I score 193 more points, 15 more scripts will become available, including several I would love to try (Runic, Hangul, Ogham, Linear B, Pheonician, Gothic, Cherokee). 643 points will unlock a further 15 scripts—of which the only one I have heard of is Miao.
The leader board shows that my 107 points put me in 557th place and that I haven’t created any unique rules. The current leader has 23,545 points and has created 563 unique rules.
Setting up a new rule
I am now setting up another rule for Zanzabar Square. Here are the 5 characters fitting the rule I am creating.
And here is my description of the rule.
The next screen shot is part way through my attempt at playing the rule. At this point, I have found 3 of the 5 characters covered by the rule.
I found the other 2 letters and passed the test. My 5 points have moved me up 25 places to 532nd.
On my first attempt (at Old Permic), I failed several times because I hadn’t checked my rules carefully enough. Either I didn’t specify the rule precisely enough or I had been careless checking which letters actually matched the rule. The app is very unforgiving. That is the point of course: the researchers want the rules to be clear and for the application of the rules to be reproducible.
One tip for playing the game if you discover that you have been imprecise or have made a mistake. At any time before your 3rd attempt, you can delete the rule. This enables you to start the rule again, selecting the letters more accurately or making your description more precise. But once you fail the 3rd time, you can no longer delete the rule.
The research team is led by Dr Yoolim Kim (Korea Institute at Harvard and Max Planck Institute for the Science of Human History) and Dr Olivier Morin (Max Planck Institute for the Science of Human History and PSL University, Paris). Here is the press release An Online Game to Crowdsource the Science of Letter Shapes | Max Planck Institute for the Science of Human History (mpg.de)
Thanks to Daniel Lazar who told me about this.