24 Stochastic Language Models 1

Purpose:

Description of the distribution and probabilities of words

Unigram models: frequency, probability of single words:

Bigram, digram models: frequency, probability in the context of one preceding word (i.e. probability of pairs):

n-gram: probability of sequences of words of length n

Estimates of prior (a priori) probabilities of words for speech recognition:

where is the acoustic model
and is the language model