I'm not exactly sure what you are rambling about, but sounds like statistical text analylis to me.

You may want to look into Text analytics and Text mining.

Also, reading about Markov chains, Bayesian statistics and Neural networks might be a good idea, since they all can be used to mathematically model different aspects of text. Good examples for this usage would be spam detection and matching anonymous text snippets to their potential authors.

What you have be aware of, though, is this aspect: Speech itself has no intrinsic value in terms of meaning, it is merely one of the methods of communicating thoughts. If you want to do more than mathematical analysis of text, you would have to find a way of associating meaning to the text snippets. Since humans always combine emotions with learnings, this you would have to replicate as well. There is an article here that describes this.