algorithm - what's the greedy or dynamic programming approach for this? -
suppose making sentences using bi-gram, means probability of appearance of each word dependent on previous word. probability of sentence multiple of probability of words
p(sentence) = p(t0)*multiple i=1 i=n p(ti|ti-1)
we have probability matrix can use determine p(ti|ti-1)
, want find probable sentence
is there greedy or dynamic programming approach it?
you can use viterbi algorithm. states words (t0, t2, t7, ...
). initial state t0
, have matrix transition probabilities a_i,j = p(tj|ti)
, have no "observations", can not think p(y|k)
. every length (t
) , every word (t_k
) find v_t,k
probability of probable sentence t
words , word t_k
@ sentence's end.
Comments
Post a Comment