This document discusses class-based n-gram models of natural language. It outlines that n-gram models predict the probability of a word based on the previous n-1 words, and that increasing n improves accuracy but decreases reliability due to more parameters needing estimation. It also discusses using word classes in n-gram models by grouping words with similar contexts together, and finding "sticky pairs" of words that occur near each other more than expected based on their individual frequencies.