## A parser for the twitter data file with all the JSON data. Reads in the JSON data and prints out a matrix of the form:

## 0, word1, word2, word3, ..., wordn

## tweet1, 0, 1, 1, ..., 0

## tweet2, 1, 1, 0, ..., 0

## ..., ..., ..., ..., ..., ...

## tweetm, 0, 1, 0, ..., 0

## where n is the number of unique and relevant words in the whole set of Tweets and m is the number of Tweets that are successfully extracted from the data.

## There are 1's whenever the word of a particular column appears in the Tweet of a particular row, else there is a 0.

## This file can then be used for Sparse PCA.

import json

from bagOfWords import BagOfWords

from dictionaryOfWords import DictionaryOfWords

from Tweet import Tweet

# To test, uncomment the next two lines

#jsonFile = open('/Users/theopavlakou/Documents/Imperial/Fourth_Year/MEng_Project/TWITTER Research/Data (100k tweets from London)/twitteranalysis/testJsons', 'r')

#commonWordsFile = open('/Users/theopavlakou/Documents/Imperial/Fourth_Year/MEng_Project/TWITTER Research/Data (100k tweets from London)/twitteranalysis/commonWords', 'r')

jsonFile = open('/Users/theopavlakou/Documents/Imperial/Fourth_Year/MEng_Project/TWITTER Research/Data (100k tweets from London)/twitteranalysis/xab', 'r')

#jsonFile = open('/Users/theopavlakou/Documents/Imperial/Fourth_Year/MEng_Project/TWITTER Research/Data (100k tweets from London)/twitteranalysis/tweets_London_22Sep12_03Oct12', 'r')

#commonWordsFile = open('/Users/theopavlakou/Documents/Imperial/Fourth_Year/MEng_Project/TWITTER Research/Data (100k tweets from London)/twitterdataanalysis/CommonWordsPlain.txt', 'r')

print("--- Loading Tweets ---")

tweetSet = []

# Make a decoder to decode the JSON string

decoder = json.JSONDecoder()

while True:

try:

# Get all lines in the file

line = jsonFile.readline()

# The lines will contain a non-empty string until the eof

# so we can break the loop when this happens

if line == "":

break

# Translate the JSON string into python JSON representation

jsonObject = decoder.decode(line)

# Make a new tweet and add it to the set of tweets that we have

# TODO

# For some reason this doesn't work when I add the date. Must fix.

#tweet = Tweet(jsonObject["text"], jsonObject["created_at"])

tweet = Tweet(jsonObject["text"])

# print("Tweet content is " + tweet.content + ", and it was created on " + "NOT TODAY")

tweetSet.append(tweet)

# Some of the lines have encoding errors so ignore them

except UnicodeDecodeError:

pass

except:

pass

print("--- Finished loading Tweets ---")

print("--- Printing size of Tweets ---")

# Print the number of tweets

print(" Size: " + str(len(tweetSet)))

print("--- Loading most common words in the Tweets ---")

# Make a list of all the unique words in the tweets which will be the columns of the matrix.

# Should reduce the > 75,000 words we have by about a factor of 10

#bagOfWords = BagOfWords(0.015)

dictOfWords = DictionaryOfWords()

for tweet in tweetSet:

dictOfWords.addFromSet(tweet.listOfWords())

#bagOfWords.addFromSetWithProbability(tweet.listOfWords())

#listOfWords = bagOfWords.getListOfWords()

#listOfWords = dictOfWords.getMostPopularWords(2500)

listOfWords = dictOfWords.getMostPopularWordsAndOccurrences(3000)

print("--- Finished loading most common words in the Tweets ---")

print("--- Printing number of unique and relevant words ---")

print(" Size: " + str(len(listOfWords)))

print("--- Finished printing number of unique and relevant words ---")

# Make a matrix of size (len(tweetSet)+1)x(len(setOfWords)+1). The first row will have the words and the first column will have the tweets (NOT the zeroth).

print("--- Opening file to output result to ---")

fileOut = open("S", "w")

# Make a file that has the words that are indexed by the columns of the matrix with their index to the left.

print("--- Opening file to output index of words to ---")

wordsFile = open("commonWordsIndex", "w")

# Matlab starts indexing from 1

i = 1

for (word, occurrence) in listOfWords:

wordsFile.write(str(i) + " " + word + " " + str(occurrence) + "\n")

i = i + 1

print("--- Closing file to output index of words to ---")

wordsFile.close()

# Make a file that has the tweet content of the tweets that are indexed by the rows of the matrix with their index to the left.

#print("--- Opening file to output index of tweets to ---")

#tweetsFile = open("tweetsIndex", "w")

# Matlab starts indexing from 1

#i = 1

#for tweet in tweetSet:

# tweetsFile.write(str(i) + " " + tweet.content + "\n")

# i = i + 1

#print("--- Closing file to output index of tweets to ---")

#tweetsFile.close()

# For each word, put a 1 for each column that has that word in the associated tweet.

# Put 1's in the matrix where the word appears and 0's where the word doesn't appear.

# Print to an output file as a .csv

print("--- Printing matrix to file ---")

# Each word is a row

for (word, occurence) in listOfWords:

# Each tweet is a column

for tweet in tweetSet:

print(word + " " + tweet.content)

tweetWordDict = tweet.dictionaryOfWords()

if word in tweetWordDict:

fileOut.write("1, ")

else:

fileOut.write("0, ")

fileOut.write("\n")

print("--- Finished printing matrix to file ---")

print("--- Closing file to output result to ---")

fileOut.close()

print("--- End ---")