Feature Engineering for Sentiment
Tuesday, November 18, 2008, 10:30am - Tuesday, November 30, 1999, 0:00am
Sentiment analysis upon free text is a difficult domain since free text is often informally written, poorly structured, and rife with spelling and grammatical errors. These characteristics make them difficult to parse and process with standard language analysis tools. These factors have made machine learning techniques such as bag of words support vector machines very popular. We describe a better feature space to use with support vector machines that relies upon the uneven distribution of words commonly seen in polar text.