So I am working on a framework called EVER/Flux, have been researching some theories for a while, and want to share this one for others to use also. In NLP, and linguistic comprehension, generally there’s somewhat of a linear approach to comprehension, for example if we take the following :
“The cat walked down the road.”
We start by assigning the to their correct parts of speech (verb, noun, etc.), but other than a basic token relationship, I found that we can do MORE. If we start to compare aspects of the sentence to one another, there’s more that can be derived because - Language is actually multidimensional
cat→down, down→walked
So we do this to derive all possible combinations, forwards, backwards, and leveraging circular/reflective reasoning when there is gain, until we have compared ALL combinations, and have a collection of comparisons. This is where is gets REALLY cool. We can do a meta-comparison between each comparison we have collected. Given this is done in scope of the linear comprehension, we derive a deeper comprehension of the phrase.
The way I am utilizing this is by having the framework have components leveraging primitives that relative to the framework correlate its comprehension of it being a verb, noun, etc. relative to the function that represents it through the primitive.
It gets REALLY COOL when you start to digest definitions and research on concepts, given you can build a relational graph.