P.s. Richard Socher, one of the authors on the paper taught a great Stanford course 'CS224d: Deep Learning for Natural Language Processing' with videos and notes available here:

http://cs224d.stanford.edu/