I wish to compare similarities/differences between human and software performance  of sentiment analysis of a specific corpus of diplomatic letters. For the time being, I am interested in the overall letter sentiment (negative, neutral, positive) . 

Similar questions and discussions