https://developer.twitter.com/ Go there to get access to tweepy

How it Works

Prep

The process begins in twitter_handler.py. The class TwitterHandler deals with all of the Twitter API interactions via python package tweepy. The only thing that is required from the user to interact with Twitter is consumer tokens and accessor tokens for a Twitter app. Those tokens should be placed in the empty strings in Main.py, and the authentication and such is handled for the user in twitter_handler.py. The function TwitterHandler.get_all_tweets() will grab all tweets from a given user under TwitterHandler.screen_name, and compile them into a single text file. Each tweet begins with <SOS> (start of string) and ends with <EOS> (end of string).

Guiding Structure

gracebot.py holds the guiding structure of the string generation algorithm. An outer-dictionary grdic holds, as a key, every non-terminating word in every tweet from the given username. Outer-keys in grdic each map to dictionaries with, as inner-keys, words following the outer keys. The inner-keys then map respectively to how frequently the inner-key follows the outer key.

Example Tweets:

1. I enjoyed our walk today
2. I thought our walk was good
3. I enjoyed today

Corresponding Example Dictionary:

{

 "I" : {"enjoyed : 2, "thought" : 1},

 "enjoyed" : {"our" : 1, "today" : 1},

 "our" : {"walk" : 2},

 "walk" : {"today" : 1, "was" : 1},

 "thought" : {"our" : 1},

 "was" : {"good" : 1}
 
 }

Example Inspection: "enjoyed" followed "I" twice, so grdic["I"]["enjoyed"]=2.

Execution

To generate a tweet, we take a walk through the dictionary defined in the previous section. The user will provide a "seed word" as an input to gracebot.generateSentence(_). The first word is used to determine the starting point of the walk, and the following words are sampled from the inner-dictionary.

Example Walk: INPUT: "walk"

• We begin at grdic["walk"] and see that there are two possible directions, "today" and "was." There is a 50/50 chance for one to be chosen over the other, since they bot follow "walk" the same amount of times. We sample grdic["walk"] and find "was."

• We are now at grdic["was"] and see that there is one possibility "good." grdic["good"] does not exist, so our output will be "walk was good."

Sampling the inner-dictionary, found in gracebot.takeLastWordAndPredict(_) will create a distribution of the inner-keys and choose an inner-key with probability inner-key's value/sum of inner-keys' values

Usage

TODO. Message me directly for questions