I thought the same few years ago – putting recognized symbol above is simple and effective. But you can get the same effect (confidence in recognized input) wit

I thought the same few years ago – putting recognized symbol above is simple and effective. But you can get the same effect (confidence in recognized input) with the Smart Script technique when done right. I briefly experimented with that in 2022 (nothing published though) and it worked pretty well. However, the implementation has to be good – you can't just beautify every symbol because it becomes distracting really fast. The less confidence system has about some symbol, the more it should fix it. High-confidence misclassifications will still be a theoretical problem, but I didn't experienced them that often.