Music and human emotion, as well as the written word, are inextricably linked. This paper aims to design a model that utilises journal-style user entries to recommend emotionally relevant music playlists. The task is completed in three parts, namely, the extraction of emotion from the text entry, the classification of a database of music by emotion, and the selection of tracks based on the emotion of the user input. Although existing research tackles each part of the task separately, the combination of the three for practical applications remains unexplored to the best of our knowledge. We describe a heuristic approach to extract emotional content from a text entry, and test various machine learning algorithms for music classification. This paper links two previously parallel aspects of affective computing as the basis for a novel way to discover music.