But it is!
Basically, NAMC is a project of mine, intended to be used to generate text that sounds like something a human would write. It didn't work. I'm putting it here since I think it's cool, and it still outperforms a normal markov chain in my opinion (for text generation).
A Markov chain is pretty simple; for each state, in this case the last word generated, there are a number of states it can choose next. For example, for the word the, the states it could choose given the sentence the brown dog was under the shed but the shed did not shelter the dog would be:
the -> brown
the -> shed
the -> shed
the -> dog
We can expand this to every word in the sentence, which results in:
-> the
the -> brown
brown -> dog
dog -> was
was -> under
under -> the
the -> shed
shed -> but
but -> the
the -> shed
shed -> did
did -> not
not -> shelter
shelter -> the
the -> dog
dog ->
Notice how there are two places where no word is listed; this indicates either the beginning of the sentence or the end. Notice how some words are mapped to multiple things; this results in something like:
-> [the]
the -> [brown, shed, shed, dog]
brown -> [dog]
dog -> [was, []]
was -> [under]
under -> [the]
shed -> [but, did]
but -> [the]
did -> [not]
not -> [shelter]
shelter -> [dog]
The initial state will be the. Now, until the state is [], we pick a random item the state can map to. The result could look something like:
the shed did not shelter the shed but the brown dog was under the dog
You can also get overly short or nonsensical results like the dog or the shed did not shelter the shed did not shelter the shed did not ... shelter the dog.
My idea was pretty simple: Rather than looking only at the previous word (current state), what if you looked multiple levels back? This is the basis of Not-A-Markov-Chain. For this example, we'll use a depth-2 NAMC. This means, rather than a single map, we get two:
D1:
-> [the]
the -> [brown, shed, shed, dog]
brown -> [dog]
dog -> [was, []]
was -> [under]
under -> [the]
shed -> [but, did]
but -> [the]
did -> [not]
not -> [shelter]
shelter -> [dog]
D2:
-> [brown]
the -> [dog, but, did, []]
brown -> [was]
dog -> [under, []]
was -> [the]
under -> [shed]
shed -> [the, not]
but -> [shed]
did -> [shelter]
not -> [the]
shelter -> [dog]
WIP