Are there any training data of tweets with labels: positive, negative, neutral as following:
"tweet 1" positive "tweet 2" positive "tweet 3" neutral "tweet 4" negative
Or in general are there any good site for finding training sets?
You can find good training datasets on kaggle!
A good example would be the Sentiment140 dataset with over 1.6 million tweets. They have been annotated with (0 = negative, 4 = positive) - but it's explained on-site! :)