Neural Networking for Dummies

So I want to write neural network that would process N-symbol words.

Should I provide 26 symbol neurons for each letter position, so that they give 1.0 for their matching letter? Should I duplicate position for each recognized words? I.e. vocabulary of 1000 N-sized words would require 1000*N*26.

>>1
It's a lot more complicated than that. A lot. It's very unlikely you'll be able to teach language to a brain whose only external stimuli are letters. For example, what is "a letter"? Your brain immediately associates it with "visual symbol" or "smallest fraction of a word". But that's only because of the phrase I wrote right before I asked you that. Had I not mentioned that meaning of the word "letter", you might have thought of an mail envelope and such.

To handle and distinguish all of those, such a brain would need to actually understand and, to a certain extent, perceive the various concepts you present to it. Anything it can't perceive or even imagine would be abstract, and abstract things are difficult to learn.