On the Github page with the public databases, one that I found epecially intersting was Sentiment140, which uses tweets for sentiment analysis. I thought it was fascinating how the model was able to find "meaning" by reading people's tweets, while taking things like sarcasm and comparisons into account. I'm curious as to whether or not using public tweets to train the model brought up any questions over rights/ownership of the data.