Sentence to Tensor

Sentence2Tensor.py

# Arda Mavi
import string
import numpy as np

characters = string.printable  # All printable characters.
token_index = dict(zip(range(0, len(characters)), characters))
max_word = 140
max_length = 80
char_len = 100 #len(token_index)

sentence_array = np.zeros((max_word, max_length, char_len))

def encode(sentence):
    this_sentence_array = np.array(sentence_array, copy=True)
    for i, word in enumerate(sentence.split(' ')):
        for j, char in enumerate(word):
            index = characters.index(char)
            this_sentence_array[i, j, index] = 1.
    return this_sentence_array

def decode(array):
    this_sentence = ''
    for i, word in enumerate(array):
        if word[0].nonzero()[0].size == 0:
            break
        for j, char_list in enumerate(word):
            char_index = char_list.nonzero()[0]
            if char_index.size == 0:
                break
            char = token_index.get(char_index[0])
            this_sentence += char
        this_sentence += ' '
    return this_sentence[:-1]