Further, I’m dubious about your classifying less well-attested natural languages as “less natural” than better-attested natural languages. The languages themselves are equally natural; it’s only our knowledge of them that differs. It would be clearer, perhaps, if you say that *our reconstruction of* a dead language is something of a conlang, generally more naturalistic than a conlang made up from scratch but less natural than a living natlang, or than the dead natlang in question was when it was spoken.

It seems to me that borrowing vocabulary from other natlangs is not a less natural process than compounding, derivation, or stretching the meanings of old words. But if you want to classify languages as less natural because they use many loanwords, why is English in category #1?

You discuss language academies etc. a little in the latter part of your post, but they don’t seem to figure in your classification scheme proper. I would suggest that colloquial spoken French is as natural as any other natlang spoken by a large and vigorous speaker community, but that standardized written French as regulated by the Academy is somewhat less natural than the typical natlang, though more natural than a reconstructed proto-language or a conlang that has no speakers.

As an earlier commenter pointed out, your classification of conlangs as more or less natural, more or less artificial, only takes into account their origin as imitating some natural language more or less closely (and that, only in a particular way). It leaves out of account whether the language has acquired a speaker community whose actual usage is now normative, rather than (or more so than) the original language inventor’s blueprints. Esperanto, Interlingua, Ido, probably Toki Pona, maybe even Lojban are all “more natural” by this criterion than Latino Sine Fleksione or Sambahsa, which you rank above them in naturalness.

Furthermore, your system takes no account, as far as I can tell, of conlangs derived from natural languages not by “correcting” them, as you describe LSF and some other auxlangs, but by putting them through a fictional but naturalistically plausible alternate history of sound changes, grammar changes, and so forth — e.g. Brithenig, Wenedyk, and so forth. I would consider these to be more naturalistic than Interlingua or Esperanto, by the origin criteron, but less natural (by the “live speaker community” criterion) than those actually spoken auxlangs. This suggests that we might want a multi-dimensional scheme for classifying languages, taking into account both size and activity of the speaker community, adherence to known linguistic universals, origin of vocabulary and grammar, etc.

What are you talking about? Esperanto is still the biggest articifial language, in terms of literature output, speakers, history etc. It has failed to take over the world, but not as a language.

I’m sorry to hear you are fed up with my opinion about Uropi not being related to Proto-Indo-European. But I think you contradict yourself by asserting Uropi is related to PIE and at the same time:

Of course, Uropi has not kept the original shape of these roots which seem rather prehistoric to me.

And giving examples of how you changed your selected vocabulary, as e.g. *sawel for “sol”, because you found it “less prehistoric” that way. Not to talk about the morphology and syntax, which are rather “modern European made easy”, as you put it.

For me – and for some others with which I’ve discussed it – yours is another artificial language, loosely based on PIE, of course, just like Esperanto was loosely based on some European languages. Which is fine.

Personally, I think it was, as an auxlang project, another step in the right direction after Esperanto’s failure. Not less, but not more than a personal conlang. Good luck.

I would attach “spoken language” to Esperanto. It is spoken in a way that Latino sine flexione or Novial are not.