3D Tensor(Typical Natural Language Processing)

• Step

1. Devide the sentences into words.

2. Vectorize each words.

3. Set up the batch size.

 

Let's say here is data.

[[나는 사과를 좋아해], [나는 바나나를 좋아해], [나는 사과를 싫어해], [나는 바나나를 싫어해]]

 

1. Devide the sentences into words.

[['나는', '사과를', '좋아해'], ['나는', '바나나를', '좋아해'], ['나는', '사과를', '싫어해'], ['나는', '바나나를', '싫어해']]

 

2. Vectorize each words.

'나는' = [0.1, 0.2, 0.9]
'사과를' = [0.3, 0.5, 0.1]
'바나나를' = [0.3, 0.5, 0.2]
'좋아해' = [0.7, 0.6, 0.5]
'싫어해' = [0.5, 0.6, 0.7]

↓

[[[0.1, 0.2, 0.9], [0.3, 0.5, 0.1], [0.7, 0.6, 0.5]],
 [[0.1, 0.2, 0.9], [0.3, 0.5, 0.2], [0.7, 0.6, 0.5]],
 [[0.1, 0.2, 0.9], [0.3, 0.5, 0.1], [0.5, 0.6, 0.7]],
 [[0.1, 0.2, 0.9], [0.3, 0.5, 0.2], [0.5, 0.6, 0.7]]]

 

3. Let's say, set the batch size as 2.

첫번째 배치 #1
[[[0.1, 0.2, 0.9], [0.3, 0.5, 0.1], [0.7, 0.6, 0.5]],
 [[0.1, 0.2, 0.9], [0.3, 0.5, 0.2], [0.7, 0.6, 0.5]]]

두번째 배치 #2
[[[0.1, 0.2, 0.9], [0.3, 0.5, 0.1], [0.5, 0.6, 0.7]],
 [[0.1, 0.2, 0.9], [0.3, 0.5, 0.2], [0.5, 0.6, 0.7]]]

 

4. Let's say, imagine it. The size of tensor is 2*3*3 (batch size, length of sentence, vector dimention)

 

vector dimension, [0.1, 0.2, 0.9] is 3.