I don't think this would be difficult unless the hand-drawn font doesn't share exact path data. If you're duplicating these characters from the same source then it wouldn't be hard to recognize paths as a specific letter:

The letter A will always have the same amount of pathPoints, the length from anchors A1 > A2 and so on is always going to be the same and even if scaled differently (since it's still proportional), so if you transpose the PathPoints to being relative instead of absolute by something like any anchor of geometricBounds then you could work with ratios and the length from A1 > A2 then A2 > A3 even if not identical would still be a multiple or proportionate from one scale to another, and the same applies to handles.

The difficulty of going this route would be that you'd need to record each character that needs to be recognized and store this in something like a JSON so that a script could parse it then take PathPoints and do a lookup to:

1.  Find every letter that has the same amount of pathPoints and same amount of handles inside as a possible match
2. Transpose the incoming letter as relative coordinates
3. Either compare all x and y positions including handles or do something like using a Math.hypot() polyfill or other simple means to just get the distance between two coordinate points and measure distances between 1 > 2 > 3 to compare each, though on false match you'd need to continue checking subsequent points and divide the smallest length by largest to get a ratio and begin checking if the ratio is the same for every point which accounts for differences in scaling, and this would need to be done on every single point and handle

If this matches then it'd be the same exact character. This is probably more complex than you're looking for but I do think it's important to point out that sure, it could be done through scripting and Javascript if you have a shared character base since you just need coordinate data.

ExtendScript alone is very difficult; AppleScript or CEP (Node.js) makes it relatively easy to achieve.

It is recommended to use OCR for text recognition. For example, the following.

• Microsoft Azure Computer Vision API
• Google Cloud Vision API
• Tesseract.js

The process flow is as follows: 

1. Convert compound path to png by imageCapture
2. Recognize text from png by OCR (AppleScript or Node.js are needed at this step)

Can you post some indicative example files? (Post as .pdfs because forum software doesn't accept .ai files.)

Hi @AliTleiss, so just to confirm: each letter path is different, for example an "C" path may have different number and placement of points than another "C"? Or will each "C" be identical (aside from position, scale and rotation of course).

- Mark

There's a slight chance these elements could be targeted. If there was a pattern a script could follow it could be possible. For example, the letter B should be a compound path with 3 path items in it, if you don't expect other art to have this particular setup then it could be found.

The plugin VectorFirstAid can do it provided that the original font is installed on your computer.