Using Bayes theoreum to train an algorithm which seeks to classify whether a tweet is more likely to be referencing 'tory' or 'evil' ... functioning on the (somewhat) biased premise that the two terms are ironically interchangable!
Pulling directly from the Twitter api calls for a great excuse to practise data wrangling with python strip functions and nltk library in order to void interferring characters and increase accuracy of classifier outcome based on words alone