This is a very interesting study--nice work! Our company does social research for all types of companies and we've found most automated sentiment analysis to be subpar (at best, 60% accurate). The problem with just looking at words is that there is no context of the whole Tweet and computers are generally bad at picking up sarcasm, innuendo and turns of phrase that may contain negative words in a positive manner (toward the brand or company).

I realize that this isn't the key focus of your paper, but we've found that sampling and human analysis/tagging is far more accurate at judging the sentiment around a brand, company or topic.