12/8/2015 0 Comments U.S. Presidential Speech AnalysisGoogle's Word2Vec is mind blowing as it converts the words to vectors which carries semantic meaning. Thus, to test the use of these word vectors I started training the word vectors on presidential speeches. After obtaining 300 dimensional vectors I accumulated word vectors to form speech vectors. It was interesting to see how the speech vectors from same president were close. I used tSNE for converting high dimensional vectors to 2 dimensions. Code repository is here: https://github.com/prateekpg2455/U.S-Presidential-Speeches Another interesting observation was to see how the speeches were different around the times of war. This is indicated by the red colored region in the speech distances matrix. Years of larger distance is around the years 1940 - 1970s.
0 Comments
Leave a Reply. |
|