I decided to play around a little with the Stanford NPL Sentiment Analysis tool. http://nlp.stanford.edu:8080/sentiment/rntnDemo.html

I used extracts from the Futurist Manifesto (1909). The screenshots below show some of the limitations of this kind of text analysis. By identifying the polarity of individual words without regard to the context or intention of the author, the results are inaccurate and largely meaningless. Those statements which would have been regarded as positive by Marinetti (the author) are labelled as negative and vice versa. Also, working at sentence level only (I couldn't get the tool to work with anything longer) is restrictive. The possibility to work at paragraph, text or concept level would be much more useful. 

Key for the trees below: orange is designated as negative and blue as positive.