Last active
October 19, 2015 16:11
-
-
Save multivac61/38d709e3be2286bffef2 to your computer and use it in GitHub Desktop.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
def _viterbi(self, observations): | |
'''Generate the most likely sequence of words given the observations''' | |
transitions = self.transition_prob; emissions = self.emission_prob | |
V = {}; V[(0, X0)] = log10(1) # initial probabilities | |
path = {}; path[X0] = [X0] # path to most likely seq | |
current_words = [[] for i in range(len(observations)+1)] | |
current_words[0] = [X0] # list of words that have an entry in bigram table | |
for ind, obs in enumerate(observations): | |
newpath = {} | |
high_prob = -INFINITY | |
filtered_words = [j for (i,j) in transitions.keys() | |
if i in current_words[ind]] | |
for w in filtered_words: | |
for q in current_words[ind]: | |
trans_prob = transitions.get( (q, w), -INFINITY ) | |
edit_prob = emissions( obs, vocab_dict.get(w) ) | |
prob = trans_prob + edit_prob + V.get( (ind, q), -INFINITY ) | |
if(prob > high_prob): | |
V[(ind+1, w)] = prob | |
newpath[w] = path.get(q) + [w] | |
high_prob = prob | |
current_words[ind+1].append(w) | |
path = newpath | |
# No we get the most likely path taken | |
filtered_keys = [(i, j) for (i,j) in V.keys() if i==ind+1] | |
filtered_dict = {key: V[key] for key in filtered_keys} | |
index, item = max(filtered_dict.iteritems(), key=operator.itemgetter(1))[0] | |
return path[item] |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment