If you do word-level markov generation, do you have to do a transition matrix of size (dictionary)X(dictionary)? Seems...inefficient.
Conversation
Replying to
you only do it for the non-0 cases: {'word': ['next1', 'next2', next3']}
1
2
Replying to
So an adjacency list rather than an adjacency matrix, still dict*M instead of dict^2... better I guess
1
Replying to
fortunately software has become very good at dealing with sparse matrices like this one.
1

