Otherwise you just get a word salad generator that produces good poems by accident. Maybe we need to try word2vec based approaches.
Otherwise you just get a word salad generator that produces good poems by accident. Maybe we need to try word2vec based approaches.
This particular sally can be found at http://sfbay-anarchists.org/wp-content/uploads/2012/05/Trurl...
What kind of grammar did you design ?
Do you think we can machine learn the grammar aided by things like word2vec ?
To be fair, it does seem like this poem almost has some nice pseudo-rhyming scheme in it
This is impressive from an AI and machine learning perspective but NLP is one area where I feel there haven't been any fundamental breakthroughs yet as we've seen for images.
Perhaps language is much more complex. Visual processing is present in many animals and so it might be that it's simpler at some level. There are very few animals that have evolved to use language like we do.
So the underlying models must be far more complex than stacking layers of convolutional layers on top of each other.
And compared to the spectacular results in images like Neural style transfer , we're judging these to a very low standard. Here is an actual poem :
“You know how this is:
if I look
at the crystal moon, at the red branch
of the slow autumn at my window,
if I touch
near the fire
the impalpable ash
or the wrinkled body of the log,
everything carries me to you,
as if everything that exists,
aromas, light, metals,
were little boats
that sail
toward those isles of yours that wait for me.” ― Pablo Neruda, The Captain's Verses