(1) vocabulary
create a dictionary consists of each unique word in the corpus
example:
I am happy; I am sad
dictionary=[I, am, happy, sad]
I am happy = [1,1,1,0]
I am sad = [1,1,0,1]
problems: large dictionary, a lot of zeros, large training and prediction time

(2) counts
still need a dictionary = [I, am, happy, sad] ---->
use this dictionary and class 1 and class 2 data to create a frequency dictionary for each class
use frequency dictionaries to generate feature vectors
[1, sum pos. frequencies, sum neg. frequencies] # 1 is for bias # sum pos. frequencies generated from instance and pos. frequency dictionary
# the key here it is 3 dimension vector, less storage, fater training
example:
class 1: [I am happy, I am happy]
class 2: [i am sad, i am sad]
class 1 frequency dictionary = [2,2,2,0] <----
class 1 frequency dictionary = [2,2,0,2] <----