The problem with the conventional way distinguishing between analytic and synthetic is defining exactly what is meant by semantic meaning:
An “analytic” sentence ... has historically been characterized as one whose truth depends upon the meanings of its constituent terms (and how they’re combined) alone... https://plato.stanford.edu/entries/analytic-synthetic/
We can make the analysis much simpler by defining the analytic / synthetic distinction from the perspective of the synthetic side.
When we define synthetic knowledge as any expression of language that requires sense data from the sense organs (Eyesight, hearing, taste, touch, smell) to verify its truth, then we have a much simpler measure.
Then analytic knowledge is what is left over: any expression of language that has been verified as true without the need of any sense data from the sense organs.
Once we eliminate sense data from the sense organs the only other way that we verify the truth of an expression of language is the semantic meaning of this expression.
Now Quine's conclusion that there is no clear distinction between analytic and synthetic is refuted by this new criteria providing the clear distinction.
There may be boundary cases that are still a little difficult to decide. The conventional way of handling these boundary cases would be to maximize the conformance to the original notions of the terms synthetic and analytic.
If we say that a particular individual human is male: Is this a definition or based on prior observation? If this is a definition then it is analytic if it is based on prior observation then it is synthetic.