Your final point is good. A graphic-to-lexicon mapping would be very useful for research and education, and I am not familiar with what is available in this area. If this is your plan, it might have very broad applications. However the mapping would be very loose because categories are never absolute (hence Sapir Whorf). Case in point: some cultures agree by shaking heads and disagree by nodding. There was a time in China where the red light meant "go" and green meant "stop." You may thus have pictures which mean exact opposites to two different people.
It's just a problem I'm pointing out, and that a YC-style voting is not going to realize the potential of the idea. But my case about the prototype is very solid though. Your definition of a chair is not "correct." Nobody's is. I happen to lay my buttox on a teapot, now it's also a chair. Not trying to cavil here, but think of how that would affect your program if you type "chair" and a teapot shows up.
It's just a problem I'm pointing out, and that a YC-style voting is not going to realize the potential of the idea. But my case about the prototype is very solid though. Your definition of a chair is not "correct." Nobody's is. I happen to lay my buttox on a teapot, now it's also a chair. Not trying to cavil here, but think of how that would affect your program if you type "chair" and a teapot shows up.