Something similar was done, but it was explicitly recognised that the purpose of lojban was not to generate "the semantic primes of language." Such as exercise is regarded by some linguists as meaningless, and by others as too difficult. Instead, concepts were listed, and from them a "covering set" was extracted. Similarly tenses, both spatial and temporal.
After the concepts were agreed, it was expressed in each of the (then) six major world languages. The words thus obtained were put through a weighting algorith,=m to try to find a "word" that had components of each, and that became the lojban word for that concept.
You can see the etymology and vocabulary here: http://en.wiktionary.org/wiki/Index:Lojban/gismu
There are documents about the process and how it was set about, but I'd have to hunt.
Regarding #2
There is a predicate "nenri" (cognate "inner") which means "x1 is inside x2". Thus
"There are dogs in the kitchen"
lo gerku cu nenmi le jupku'a
lo gerku cu zvati le jupku'a
There is a web page that will gloss grammatically correct lojban here:
http://www.lojban.org/jboski/index.php
I'm working on a tool to help create grammatically correct lojban interactively.
Hope that helps.