nlp - startProbability in hidden markov models -

nlp - startProbability in hidden markov models -

how can calculate startprobabilitise in hidden markov model pos tagging? means: (# repeats of 1 tag)/(# repeats of tags)? or (number of sentences origin 1 tag)/(all sentences)? if not, what?

considering sec order hmm, maximum likelihood estimate gives:

p(sometag | <bos>,<bos>) = count(<bos>,<bos>,sometag) / count(<bos>,<bos>)

it corresponds sec proposal:

(number of sentences origin 1 tag)/(all sentences)

nlp tagging text-mining hidden-markov-models pos-tagging