def get_top_words(wc, train_sample_size, top_num = 10):
top_words = {}
for word in wc:
if not word in stopwords.words():
for decade in wc[word]:
p = P2.p_x_given_y(wc, word, decade, train_sample_size)
p_min = 1
for n_decade in wc[word]:
if not n_decade == decade:
p_min = min(P2.p_x_given_y(wc, word, n_decade, train_sample_size), p_min)
ratio = p/p_min
# ratio = wc[word][decade]/min(wc[word].values())
if decade in top_words:
w = top_words[decade]
if len(w) < top_num:
top_words[decade][word] = ratio
else:
mw = min(w, key=w.get)
if ratio > w[mw]:
del top_words[decade][mw]
top_words[decade][word] = ratio
else:
top_words[decade] = {}
top_words[decade][word] = wc[word][decade]
return top_words
