def build_lemma_vocab(append_string=''):
"""
Builds entire vocabulary from a dataframe of reviews.
"""
import process_text
# Restore data from file
reviews = pd.read_pickle('../data/pandas/review'+append_string+'.pkl')
vocab = {}
print len(reviews)
count = 0
for index,row in reviews.iterrows():
lemmas = process_text.text2lemmas(row.text)
for lemma in vocab:
try:
vocab[lemma] += 1
except:
vocab[lemma] = 1
count += 1
if count % 100==0: print count
# Save vocabulary to file
vocab = pd.Series(vocab)
vocab.to_pickle('../data/pandas/vocab'+append_string+'.pkl')
return True
def add_lemmas2pandas(type_string='', append_string=''):
"""
Adds lemmatized text as an extra column in the reviews and/or sentences
databases.
type_string = {'review'|'sentences'} determines whether to lemmatize
text in pandas dataframe of reviews or sentences
"""
import process_text
# Error handling
if type_string not in ['review','sentences']:
print "Error in add_lemmas2pandas:"
print " type_string must be either 'review' or 'sentences'"
print " please try again"
return None
# Lemmatize each row in the dataframe
dataframe = pd.read_pickle('../data/pandas/'+type_string+append_string+'.pkl')
lemmatized_text = []
count = 0
for item in dataframe.index:
#if count<63400:
# count += 1
# continue
if count%1000==0: print count
thisitem = dataframe.loc[item]
lemmatized_text.append( process_text.text2lemmas(thisitem.text) )
count += 1
# Add lemmatized text back as a column in the dataframe
lemmatized_text = pd.Series(lemmatized_text,index=dataframe.index)
dataframe['lemmas'] = lemmatized_text
# Save dataframe to file
dataframe.to_pickle('../data/pandas/'+type_string+'_lemmas'+append_string+'.pkl')
return True
