Word similarity graph


Process followed:

  1. WORD2VECTOR:  - a word2vec is trained on all and only speeches belonging to the topic 
  2. INSERTING TOP NMF WORDS:  - the top NMF words are inserted into the graph 
  3. 1ST LEVEL SIMILARITY WORDS:  - include all the words similar to each NMF word according to word2vect trained at step 1. Connect similar words with a line 
  4. n-th LEVEL SIMILARITY WORDS:  -insert into the graph all words similar to each words introduced in the (n-1)th level of iteration. Connect similar words with a line 
  5. STOP RECURISON:  - repeat recusively until the graph includes a minimum number of words