What do you Tweet? An Analysis of Twitter using Support Vector Machines
Richard Sung · Trinity College Digital Repository (Trinity College, Hartford CT) · 2012
In the past few years, Twitter has become a major social networking service with over 200 million tweets made every day. With this newfound source of expanding information, can people stay up to date with what others are posting? Along with the increasing processing power of computers, is there a way computing can analyze tweets on a large scale? Moreover, can computers understand what people think based on what they post? This senior project explores this question by determining the positive or negative sentiment of twitter posts by using a machine learning algorithm called Support Vector Machines. Based on a labeled dataset of tweets, a parser then extracts present features in the text to create a vector. Once a collection of vectors is compiled, data is trained and tested to create a working model, which can then be evaluated to determine the effectiveness of the classifier. Based on a dataset of 359 tweets and 329 features, a model can accurately classify tweets as high as 74.84% using a linear classifier.