NLP processing

 # Transforming the reviews data by removing stopwords, using regular expressions module to accept only letters,
 # tokenizing those words and then making all the words lower case for consistency.
 comments = []
stop_words = set(stopwords.words('portuguese'))


for words in review_data['review_comment_message']:
    only_letters = re.sub("[^a-zA-Z]", " ",words)
    tokens = nltk.word_tokenize(only_letters) #tokenize the sentences
    lower_case = [l.lower() for l in tokens] #convert all letters to lower case
    filtered_result = list(filter(lambda l: l not in stop_words, lower_case)) #Remove stopwords from the comments
    comments.append(' '.join(filtered_result))