Multi-Dimensional Sentiment Analysis with Learned Representations

Treating sentiment analysis as a classification problem has proven extremely useful, but it misses the blended, continuous nature of sentiment expression in natural language. Using data from the Experience Project, we study texts as distributions over sentiment categories. Analysis of the document collection shows the texts contain blended sentiment information substantially different from a categorization view of sentiment. We introduce a statistical vector-space model that learns from distributions over emotive categories, in addition to capturing basic semantic information in an unsupervised fashion. Our model outperforms several baselines in predicting sentiment distributions given only the text of a document.