The 22 categories of words used in Carl Darling Buck's "A Dictionary of Selected Synonyms in the Principal Indo-European Languages" (1949) are quite different from for instance the categories in Roget's thesaurus as of 1911. There is no background in the book itself on how the 22 categories where chosen. Did Buck make the list himself or was he following some (undocumented?) tradition?
I haven't read his Introduction for a while, but that's where I'd start. If he doesn't deal with it there, I can't help you,
however, you should be aware
- that these are all pretty standard categories in natural meaning categorizations, like PIE roots, Levin's verb lists, classifier systems, phonosemantics, etc.
- that no linguist I know is very concerned about where Buck got his 22 categories. 22 is not an unreasonable number: if it works, use it; if it doesn't, ignore it. Certainly one can use more or fewer categories; there's nothing sacred about 22 the way there is about, say, 42.