I've been thinking of putting together some sort of language, where the letters in the words are basically descriptive.Stuff like that is hard and there's no shortage of attempts of trying to categorize everything.
I'm not really sure how possible it is.
It seems really challenging to try to figure out categories, but I've only just started thinking about it.
As an example:
The first letter might describe if it's a living noun, or a verb or something. Say 'a' for a living thing.
The second letter might describe how many legs the living thing has. Say "b" means 2.
So the word "ab" would mean bipedal, and the word for human would start out with "ab".
I want to try to see if there are some good categories so that the words wouldn't become overly long.
Lots of conlangs try to have as few root words as possible and then build everything by compounding them. Toki Pona is one example.
Yours is different since the second "root" is contextual (and you want to use a single letter so you have less options) but that might give you a direction to start with.
Also maybe look into logban. Verbs in that language have very specific and defined noun relations so maybe you could do something like that. IE "A" means it's living and then that defines that the second letter is leg count, third letter is kingdom, etc.
