9 Stochastic Modelling

Main features:

1. Data collection and analysis
2. Description of the distribution and probabilities of sounds, words
3. Unigram models: frequency, probability of single words:

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

5. n-gram: probability of sequences of words of length n
6. Estimates of prior (a priori) probabilities of words for speech recognition:
   

   where is the acoustic model
   and is the language model