Most languages take meaningless sounds, combine them to create roots and affixes. The roots tend to be content, the affixes tend to be grammatical. The difference is in how specific they are. Grammatical affixes tend to be bleached of semantic meaning, compare walrus and "happened in the past".
Many categorical or taxonomic conlangs attempt to assign meaning to the phonemes and then generate words that somehow have meanings inferable from their spelling. I won't go into that approach much here because natural languages don't work that way, so there is less science available to give advice.
Morphemes tend to be different depending on the lexical category. For example, most languages with affixes allow the set of verbal affixes to attach to verbs, but not nouns, or if they do, they don't mean the same thing.
Morphemes can vary depending on what the root looks like. For example, roots ending in consonants are likely to have an V or VC suffix, while roots ending in vowels are likely to have C or VC suffixes, or a legal consonant cluster.
Representing Lexical Class and Affixes
... out of time... Check back soon, or head over to the forums..
Hint, it involves Branch, which tells the wordo engine to create multiple forms of the same word.