--- On Wed, 1/14/09, Erbrice <[log in to unmask]> wrote:

> yes but how do you split "chair", is it a concept

My own random thoughts:

The decision to call a given word a semantic prime is somewhat arbitrary, and depends on the purpose to which derivations will be put. Thus "dog" might be considered to be a primitive concept, since dogs are familiar in virtually all human populations. 

It could also be claimed that "dog" can be factored into "canine animal" except that "canine" intrinsically implies "animal" so the explicit mention of the word "animal" is redundant, and we discover that "canine [n]" is really just a synonym for "dog [n]". We might also decide to define "dog" in terms of a narrative description that uses only words from some specified set of semantic primitives as done in "Semantics: Primes and Universals" (Wierzbicka, 1996). This results in some very long and convoluted definitions that lack transparency.

For use as a practical language building tool, a prime would be more like a word that might be found in a Swadish list, and a derivation, or factorization of a composite word would be expressible using two or three factor words, at most. Thus the primes would be more like axioms in Euclid than like actual primes in number theory.

Consider the concept "young" in:

	puppy = {young dog}

We might define "young" as "recently created", although the word "created" might not be applicable in a particular instance. We might prefer "recently made", unless the young thing in question is not something that is made. Thus, what we really need is a word that means "made or created or having come into existence" We can consult the thesaurus and try other words until we hit upon one that seems to capture the essence. For example:

	young = {recently emergent}
	young = {recently manifest}
	young = {recently instantiated}

If we choose, for example, {recently manifest}, then we could also factor "puppy" as:

	puppy = {{recently manifest} dog}

The reason why this is not the best choice is that the concept "young" is itself so useful that we are better off defining "young" and then, once defined, using it.

We can also decompose "recent" as "near past", where "past" might be defined as "before now":

	past = {before now}
	recent(ly) = {near past}
	young = {recently manifest}
	puppy = {young dog}

Which could be combined into:

	puppy = {{{near {before now}} manifest} dog}

but clearly the stepwise definitions are easier to follow.