Semantics

All concepts are "fuzzy sets", that is, they are a collection of "prototypes" (most-typical examples) with some degree of variation permitted within, between, and outside of the prototypes. Even a named individual, as a concept, is a fuzzy collection of incidents (e.g. various positions and movements, appearance at different times and places, etc).

An animal, when it comes across a situation that has caused him pain in the past, will associate the concepts of that situation with the concepts of actions that have helped in the past. The results of this association will be strengthened if the results are satisfactory, or not if not. In this way, concepts become interconnected in the animal's memory to form a semantic network. But there are no words.

In humans, concepts are connected to lexemes, which operate a bit like a "fuzzy word-set". Commonly, there are one or two prototypical lexemes that are then capable of being altered by syntax and other processes which are applied to the lexemes in context with other concepts and their associated lexemes.

At first, a child learns to associate specific lexemes with specific concepts. The child then learns to add to individual lexemes a variety of details and begins to compose the resulting set of lexemes into a structure that conveys the inter-relationship of the details that is more-or-less isomorphic with the relationship of the concepts. This is semantic structure.

Another way to look at semantic structure, on a word-by-word basis, is componential analysis. A brief introduction is available here.

When someone experiences an event, a portion of the semantic net is engaged, both the concepts and the associated lexemes. While the relevant concepts and lexemes are engaged, they form a temporary structure. The conceptual structure may become a memory of the event. The lexemic structure may become the basis for a verbal expression.

(Although this temporary structure is actually within the larger net of semantic knowledge, it may be convenient to think of it as a network of lexemes in a "semantic sketch-pad, like cognitivists talk about the phonological loop and the visual sketch-pad, both of which are likely also embedded in the larger network and not limited to a localized area of the mind or brain.)

We can model this lexemic structure in various way. I prefer the model developed by Leech in his book "Semantics". Examples of Leech's model can be viewed here.

This structure is then processed by the learned syntax of one's particular language into something we can model with the familiar syntactic "trees" we find in any introduction to syntax. This in turn is further processed until it becomes a verbal expression. Of course, the process also happens in reverse (from hearing to conceptual reconstruction) and the expression portion of the process can include written as well as verbal materials.

More detail on syntax and morphology can be found here and here, respectively.