Some observations: 1) If we view N gram counts as observation counts, we can generalize Norman Megill's estimation of Bernouilli probabilities to manysided dic

Some observations: 1) If we view N gram counts as observation counts, we can generalize Norman Megill's estimation of Bernouilli probabilities to manysided dice (as many sides as token alphabet): (O+1)/(T+S) where O is the number of Occurences, T is the number of Trials and S is the Size of the alphabet or number of Sides. Having 0 observations does not result in probability of 0. 2) Since the method allows for collecting occurences in the corpus, this could be used to provide alternative language models contexts in the corpus.