def preProBuildWordVocab(sentence_iterator, word_count_threshold = 5):
# count up all word counts so that we can threshold
# this shouldnt be too expensive of an operation
print 'preprocessing word counts and creating vocab based on word count threshold %d' % (word_count_threshold, )
t0 = time.time()
total_sent_len = len(sentence_iterator)
total_world_cnt = 0
word_counts = {}
for sent in sentence_iterator:
line_seg = conv_util.segment_sentence_cn(sent)
for w in line_seg:
word_counts[w] = word_counts.get(w, 0) + 1
for w in word_counts.keys():
total_world_cnt = total_world_cnt + word_counts[w]
vocab = [w for w in word_counts if word_counts[w] >= word_count_threshold]
print 'filtered words from %d to %d in %.2fs' % (len(word_counts), len(vocab), time.time() - t0)
# with K distinct words:
# - there are K+1 possible inputs (START token and all the words)
# - there are K+1 possible outputs (END token and all the words)
# we use ixtoword to take predicted indeces and map them to words for output visualization
# we use wordtoix to take raw words and get their index in word vector matrix
ixtoword = {}
#ixtoword[0] = '.' # period at the end of the sentence. make first dimension be end token
wordtoix = {}
#wordtoix['#START#'] = 0 # make first vector be the start token
ix = 1
for w in vocab:
wordtoix[w] = ix
ixtoword[ix] = w
ix += 1
misc = {}
misc['wordtoix'] = wordtoix
misc['ixtoword'] = ixtoword
misc['total_sent_len'] = total_sent_len
misc['total_world_cnt'] = total_world_cnt
misc['avg_snt_len'] = float(total_world_cnt)/total_sent_len
print (total_sent_len,total_world_cnt,misc['avg_snt_len'])
print 'vocab size = %d'%len(misc['ixtoword'])
return misc
ix_label_out_file = nlpcaffe_data_out_dir + file_name
ix_label_out = []
for i in xrange(0,len(label_list)):
labels = label_list[i]
label_split = labels.split('\t')
if len(label_split)<2:
ix_label_out.append("")
continue
#!!!
each_label = label_split[1]
token_label = conv_util.segment_sentence_cn(each_label)
ix_token_label = []
for each_token in token_label:
if each_token in wordtoix:
ix_token_label.append(str(wordtoix[each_token]))
ix_token_label_str = " ".join(ix_token_label)
ix_label_out.append(ix_token_label_str)
print "set %s size is %d"%(file_name,len(ix_label_out))
common_io.write_txt_lines(ix_label_out_file,ix_label_out)
