About make_batch of NNLM

```python
input = [word_dict[n] for n in word[:-1]] # create (1~n-1) as input
target = [word_dict[word[-1]]]
```

it constraints the length of input and n_step. I think the following example is even better

```python
for i in range(len(words) - window_size + 1):`
` x_train.append(words[i: i + window_size - 1])`
`y_train.append(words[i + window_size - 1])
```

Start by locating the make_batch entry point in the NNLM tutorial and read how input, target, word, n_step, and window_size are currently used. Compare its generated training pairs with the proposed sliding-window example; done means the batching behavior supports the intended window length without constraining input and n_step.

Written by the indexing model from the issue text.