Determining thresholds for transitional probabilities

[Popcorn Sutton]

Hey all, I'm working on a project right now and Im going to skip right to the point.

 

I count occurrences for probabilistic training.

Sometimes my program generates a string like this 'eres your stuff'

 

Should I leave it the way it is, which is approximately 80% grammatical, or should I prettify it? Should I even waste my time on prettifying its output?

 

Another question is this. If I make a variable for strength that counts every sequence of occurrences, certain sequences are going to emerge as more probable, namely, sequences of 1 or 2 occurrences. I don't want to arbitrarily assign a number, but I'm thinking that if I add one occurrence, the most probable one, I might need to assign a 50% likelihood or so of accepting the extra occurrence.

 

I have an idea of what I should do, but I want to get your opinions before I show the maths I have in mind.

 

I answered my question in the post so sorry you guys can delete this message if you want.