3 ms·
This is quite nice. The generalization of this to full formulas is much more difficult. In most languages, letters are designed to be distinct, and in case of c
by abdullahkhalids 3y ago
This is quite nice. The generalization of this to full formulas is much more difficult. In most languages, letters are designed to be distinct, and in case of confusion, many times a secondary pass with a dictionary [1] helps quite a lot.
But mathematics symbols are often quite close, as evidenced by this app. There are multiple suggestions for each input, and they are often quite close that without larger context it is difficult to tell which one the user meant. My handwritten mathematical x, greek chi, and the times symbol look very similar. Alpha and "proportional to". Etc etc. At minimum, the ai would likely need some understanding of the structure of mathematical statements and equations to do this better.
[1] Or even a third pass with a grammar checker.
- amai 3y agoSee https://facebookresearch.github.io/nougat/ https://facebookresearch.github.io/nougat/ and https://lukas-blecher.github.io/LaTeX-OCR/ https://lukas-blecher.github.io/LaTeX-OCR/
- abdullahkhalids 3y agoThese are on digital documents, where the characters do look different. Handwritten characters look similar. It's a much more difficult task.