Does anyone know of any libraries I could use to convert a word into a semantic prime, given a list of approved primes?
In other words, given a sentence or a word, I would like to find the combination of "prime" words that best match the input semantically.
A good example of this would be a function that takes in a Wikipedia article and outputs something like Simple Wikipedia - it has the same semantic content, but restricts word usage to a number of prime words (in this case, simpler words).
Another example: say my primes are ["person", "move", "tool", "nice", "building"]
.
Then I would like to have some function, library, or NLP algorithm that converts "hammer"
into ["building", "tool"]
, "car"
into ["moving", "tool"]
, and "friend"
into ["nice", "person"]
.
My first intuition was to use word2vec: loop through every combination of the primes and find the combination of primes who's vector most closely matches the original word. However, I suspect this will not work out very well (partially because vector addition doesn't seem to be accurate enough). Is there a more robust way to do this?