Sample project for using stylometry to deanonymize Twitter account author.

    $ pip install tweepy numpy unidecode nltk scipy sklearn
    $ python3

In Python, import nltk and download Model punkt. ​

    >>> import nltk
    >>> nltk.download()

    $ git clone https://github.com/ViliamV/stylometry.git
    $ cd stylometry/

• Follow these steps.
• Input credentials into twitter-API.txt

• Create accounts.txt in main directory and put there account's names to download, one in each line. Put the unknown author's account last.
• Create directory data in main directory.
• Run tweet-downloader.py and wait. Due to Twitter API speed, it might take a while.
• Verify if data contains downloaded tweets.

• Edit classification.py and change value UNKNOWN (line 28) to unknown author's account.

    UNKNOWN="example_account"

• Run classification.py.

This code uses Bag of Words model for extracting features from the text. A great introduction for implementing this model can be found here.

The code also uses Czech stopwords and Czech tokenizer, however, it is quite simple to change it.