Before we can create a model of our taste in news articles, we need training data. This training data will be fed into our model in order to teach it to discriminate between the articles we'd be interested in and those we would not. To build this corpus, we will need to annotate a large number of articles to correspond to these interests. We'll label each article either y or n, indicating whether it is the type of article we would want to have sent to us in our daily digest or not.

To simplify this process, we'll use the Pocket app. Pocket is an application that allows you to save stories to read later. You simply install the browser extension, and then click on the Pocket icon in your browser's toolbar when you wish to save a story. The article is saved to your personal repository. One of the great features of Pocket for our purposes is the ability to save the article with a tag of your choosing. We'll use this to mark interesting articles as y and non-interesting articles as n.