How to one-hot-encode sentences at the character level?

Question

I would like to convert a sentence to an array of one-hot vector. These vector would be the one-hot representation of the alphabet. It would look like the following:

Answer 1

This is a common task in Recurrent Neural Networks and there's a specific function just for this purpose in tensorflow, if you'd like to use it.

alphabets = {'a' : 0, 'b': 1, 'c':2, 'd':3, 'e':4, 'f':5, 'g':6, 'h':7, 'i':8, 'j':9, 'k':10, 'l':11, 'm':12, 'n':13, 'o':14}

idxs = [alphabets[ch] for ch in 'hello']
print(idxs)
# [7, 4, 11, 11, 14]

# @divakar's approach
idxs = np.fromstring("hello",dtype=np.uint8)-97

# or for more clear understanding, use:
idxs = np.fromstring('hello', dtype=np.uint8) - ord('a')

one_hot = tf.one_hot(idxs, 26, dtype=tf.uint8)
sess = tf.InteractiveSession()

In [15]: one_hot.eval()
Out[15]: 
array([[0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
       [0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]], dtype=uint8)