As for that infinite regress that you talk about, all our words come from somewhere in our experience and should be able to be referred back to that experience if necessary. Hume argues if a word's foundation can't be established by an impression gained during an experience, or a copy of an impression gained from someone else's experience, then the word or idea has no validity. So there wouldn't be an infinite regression because while words require more words in order to be explained, all of the words in use should be from a foundation of possible experiences. He's mainly talking about things like "innate ideas" which Descartes used to try and prove the existence of God, but I can see how things like "natural" vs "unnatural" can be subject to this kind of criticism. Where in your experience have you seen an "unnatural object"? Isn't a scientist in a lab tinkering with his tools of chemistry acting no more naturally than an otter breaking up whatever they eat with a rock on their belly?
I am all for defining as many words as necessary to get everyone on the same page, if such a rigorous debate was for some reason necessary, it'd be the only way to make sure people are talking with purpose and meaning, and not establishing statements that have no solid basis.
Again, I'm not talking about just talking just for the sake of talking, or other forms of communication. Obviously if a method of communication has survived to this day, it must have served some purpose over the eons, so it's all useful to some degree, and in some circumstances.