seventh.py

"""
To evaluate the good or bad score of a tweet, we first tokenize the tweet, and then 
stemmize each word in our tweet. We also associate each stem with positive and negative values, 
respectively, using a dictionary.

Finally, we caculate the average word weight of a tweet, and decide if it's a good or bad one 
based on that.

"""
import json
import html
import twitter
import time

from nltk import word_tokenize, pos_tag
from nltk.stem.porter import *

# Break down a string into words
def get_words(str):
  useful_pos = {'NN'}
  tokens = word_tokenize(str)
  tags = pos_tag(tokens)
  return [word for word, pos in tags if pos in useful_pos] 

# Load a json object from a file
def load_json(json_file):
  with open(json_file) as f:
    return json.load(f)

# Calculate the average value of words in list_of_words
def get_average_word_weight(list_of_words, word_weights):
  number_of_words = len(list_of_words)
  sum_of_word_weights = 0.0
  print (number_of_words)
  if number_of_words == 0:
    return 0.0
  stemmer = PorterStemmer()
  # Iterate through the words in the tweet string
  for w in list_of_words:
      stemmed_word = stemmer.stem(w)
      if stemmed_word in word_weights:
          sum_of_word_weights += word_weights[stemmed_word]
      #else:
          #missing_words[stemmed_word] = 0.0

  return sum_of_word_weights / number_of_words

# Analyse a tweet using a word-weight dictionary
def anaylse_tweet(tweet_string, word_weights):
  words = get_words(tweet_string)
  avg_tweet_weight = get_average_word_weight(words, word_weights)
  print (tweet_string + ":" + str(avg_tweet_weight))

# Load word weights and credentials from json files
word_weights = load_json("word_weights.json")
credentials = load_json(".cred.json")

# Connect to the twitter api
twitter_api = twitter.Api(consumer_key=credentials["consumer_key"],
                          consumer_secret=credentials["consumer_secret"],
                          access_token_key=credentials["access_token_key"],
                          access_token_secret=credentials["access_token_secret"],
                          tweet_mode='extended')

# Load last 10 statuses of Donald Trump
statuses = twitter_api.GetUserTimeline(screen_name="realDonaldTrump", count=10)

# Iterate through them and analyse them
for status in statuses:
  anaylse_tweet(html.unescape(status.full_text), word_weights)