Let's put all we've learned to good use and do some data exploration for our spam filter. We will use the Ling-Spam email dataset: http://csmining.org/index.php/ling-spam-datasets.html. The dataset contains 2412 ham emails and 481 spam emails, all of which were received by a mailing list on linguistics. We will extract the words that are most informative of whether an email is spam or ham.
The first steps in any natural language processing workflow are to remove stop words and lemmatization. Removing stop words involves filtering very common words such as the, this and so on. Lemmatization involves replacing different forms of the same word with a canonical form: both colors and color would be mapped to color, and organize, organizing and organizes would be mapped to organize. Removing stop words and lemmatization is very challenging, and beyond the scope of this book (if you do need to remove stop words and lemmatize a dataset, your go-to tool should be the Stanford NLP toolkit...
