Another type of layer that
you can use is a convolution, in a very similar way to
what you did with images. The code to use a convolutional
on network is here. It's very similar to what you had before. You specify the number of convolutions
that you want to learn, their size, and their activation function. The effect of this will then be the same. Now words will be grouped into
the size of the filter in this case 5. And convolutions will learned that
can map the word classification to the desired output. If we train with the convolutions now,
we will see that our accuracy does even better than before with close
to about 100% on training and around 80% on validation. But as before,
our loss increases in the validation set, indicating potential overfilling. As I have a super simple network here,
it's not surprising, and it will take some experimentation
with different combinations of conversational layers
to improve on this. If we go back to the model and
explore the parameters, we'll see that we have 128
filters each for 5 words. And an exploration of the model
will show these dimensions. As the size of the input was 120 words,
and a filter that is 5 words long will shave off 2 words from
the front and back, leaving us with 116. The 128 filters that we
specified will show up here as part of the convolutional layer.