Pull Twitter feeds and score sentiment of tweets; then use training set to classify tweets. Also visualize the frequency of unigrams, bigrams and other ngrams, as well as stemming & lemmatization effects.

Next steps include using the features (either unigrams, or ngrams) to train the data and then classify. Additionally, analyze the retweet count as a response variable while using features (presence of terms) in text corpus to predict the probability of retweet count.

Long term steps include incorporating other social media APIs through Python (Pinterest, Facebook, Yelp, and Google+) to indicate the overall web sentiment of particular food/restaurant/business.

Also long term, analyze Uber geographically, to determine if using Census data (on demographics) shows any patterns in positive sentiment. Do older/younger, more affluent/poor, or those without as many public transportations/restaurants have varying levels of Uber sentiment?