POC: is it possible to build a meaningful language model which represents political persuasion?

[timini]

• Do individual people cluster into their political parties?
  • get party membership data for all speakers
  • create vector representations of speeches
  • cluster speeches using PCA and label by party http://scikit-learn.org/stable/modules/decomposition.html#decompositions
• Do speeches / people from the house of lords vs the house of commons cluster separately?
  • PCA of all speeches, labeled by house.
• What are the significant terms for the house of lords vs commons?
• What topics do one house prefer to talk about over the other?
  • generate the topics for house of lords
  • generate the topics for the house of commons
• What topics do each party prefer to talk about?
• Scrape news articles and try to classify using our trained classifiers

[timini]

Use this to read the matrix into scipy
http://docs.scipy.org/doc/scipy/reference/generated/scipy.io.mmread.html#scipy.io.mmread