pmonks / fake-tweets

Fake tweets generated by a Markov chain built from a Twitter archive

fake-tweets

A small library that builds a Markov chain from a tweet archive in JSON format, then spits out any number of statistically similar, though fake, tweets.

Trying it Out

Providing a JSON Archive

Download a JSON tweet archive, for example via Twitter's export function, using something like this script, or from this archive of hot, toxic garbage.

The JSON must be an array of objects, with each object containing either a text or full_text key, whose value is the full text of that tweet. No other keys are used.

e.g.

[
  {
    "source": "Twitter for iPhone",
    "text": "Who can figure out the true meaning of \"covfefe\" ???  Enjoy!",
    "created_at": "Wed May 31 10:09:22 +0000 2017",
    "retweet_count": 85555,
    "favorite_count": 209543,
    "is_retweet": false,
    "id_str": "869858333477523458"
  },
  {
    "source": "Twitter for iPhone",
    "text": "Despite the constant negative press covfefe",
    "created_at": "Wed May 31 04:06:25 +0000 2017",
    "retweet_count": 127507,
    "favorite_count": 162788,
    "is_retweet": false,
    "id_str": "869766994899468288"
  }
]

Running the Code

Clone this repo, symlink your tweet archive, then start a REPL using the provided init script:

$ git clone https://github.com/pmonks/fake-tweets.git
$ cd fake-tweets
$ ln -s /path/to/your/tweet-archive.json hot-toxic-garbage.json
$ clj -i init.clj -r

This will load and analyse the given tweet archive, then generate one fake tweet up to 100 "words" in length. To generate more:

(ft/fake-tweet markov-chain)

In addition to the markov-chain var, the init script also initialises vars named vocabulary, vocab-freq and sorted-vocab-freq which allow primitive analysis of the tweeter's vocabulary using standard Clojure functions:

; How bigly is the tweeter's vocabulary?  Note: punctuation, numbers, emojis, hashtags, @mentions etc. are all included in the count
(count vocabulary)

; Does the tweeter excrete covfefe?
(not (empty? (filter #(s/includes? (s/lower-case %) "covfefe") vocabulary)))   ; Not idiomatic Clojure, but I don't agree with the rationale...

; How much?
(get vocab-freq "covfefe")

; What are the tweeter's favourite 100 words?
(take 100 sorted-vocab-freq)

Contributor Information

GitHub project

Bug Tracker

License

Copyright © 2020 Peter Monks Some Rights Reserved

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

SPDX-License-Identifier: CC-BY-NC-SA-4.0