Tweet Sentiment Learning

Dictionary
164242       63.825 (Positive word +3, Negative +1. Penalty if "not/don't/can't" appears. Bonus if "very/so/really" appears. Bonus if emoticons present.)
155202       62.75 (Vanilla, but great punctuation removal. Removes @names and URLs entirely.)
152490       62.650
162166       62.525
152196       61.125
163468       60.975
165364       60.575
153762       60.5
155748       60.5
161518       60.425
155148       60.400
150060       60.375
150876       60.375
156714       60.3
161398       60.3
160834       60.275
164332       60.225
164818       60.175
153264       60.175
162520       59.05
157362       59.05
164488       58.425
160384       56.2
160468       56.125
161998       53.275
163456       47.55

Learned Words
155202       58.8 (Conditional probabilities, compare pos vs neg, saves if 2x as big)
162520       58.025 (Conditional probabilities, compare pos vs neg, saves if diff is big, also minimum count cutoff based on # tweets)
156714       56.725 (Conditional probabilities, compare pos vs neg by ratio)
153762       54.950 (Raw counts. Large common word list. Weighted best learned words higher in classification.)
163468       54.675 (Ratio of pos/neg frequencies, multiplied by ratio of total neg/pos tweets, also minimum count cutoff)
150876       51.150 (PMI of pos vs neg)
162166       51.025 (PMI of pos vs neg, plus other thresholds)
161998       50.55 (Raw count comparison, + full english dictionary)
157362       49.6 (Raw counts)
164242       49.05
160468       48.425
153264       48.35
164488       48.225
164332       47.5
165364       47.475
150060       47.450
160384       47.05
155148       45.0
161398       43.95
152490       43.825
161518       40.75
164818       36.35
152196       31.25
155748       31.025
160834       --
163456       --