Proof that Lojban morphology is unambiguous

[scope845h...@icebubble.org]

Does anybody have access to a proof, or know how to prove, that Lojban's
morphology is unambiguous?

I was corresponding with lojbab, for a while, about this and he assured
me that the unambiguity of Lojban's morphology had been proven
analytically. But when I asked him for his analytical methods/proof, he
stopped responding to me. :(

Does anybody know where this information can be found? It would be much
appreciated.

[scope845h...@icebubble.org]

Is it taboo to question the non-ambiguity of the Lojban morphology? I
can't seem to get a straight answer from anyone about this.

[Jacob Thomas Errington]

I don't believe there is a proof out there, but it is by no means
taboo to question whether Lojban actually achieves its stated goals.

To prove that the morphology is unambiguous, one would have to write
an algorithm that accepts an idealized representation of a speech
stream and decomposes it into a sequence of words. The algorithm would
not be able to consult a dictionary to do so.

If you're interested in working on this, I'd be interested in hearing
about it.

mu'o mi'e la tsani

> --
> You received this message because you are subscribed to the Google Groups "lojban" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to lojban+un...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/lojban/865z9z7dyr.fsf%40cmarib.ramside.

--
Jacob Thomas Errington
W: https://jerrington.me/

[Bob LeChevalier]

It is not taboo, but there isn't a lot of interest in the issue, and I
don't have time to do the digging you require, and I am not sure that
there was even a final writeup. Most of the people who were interested
in the past are inactive at this point.

This was a question that largely was satisfied over 25 years ago. We
satisfied ourselves to a level that met our needs at the time, and no
one has come up with counterexamples. The few problem areas, if they
can be called such, had to do with possible extensions to the language,
none of which are even being used to my knowledge.

I do recall that at one time we attempted to write a software exhaustive
test something like what was suggested by Mr Errington, but we decided
that the number of possible cases was far too large and hence
non-computable, given the PC hardware then available, and went to a more
logico-analytical approach. But any notes that I might have kept are
rather deeply buried in my archived files.

lojbab

[Bob LeChevalier]

On 8/3/2020 4:04 PM, Jacob Thomas Errington wrote:
> I don't believe there is a proof out there, but it is by no means
> taboo to question whether Lojban actually achieves its stated goals.
>
> To prove that the morphology is unambiguous, one would have to write
> an algorithm that accepts an idealized representation of a speech
> stream and decomposes it into a sequence of words. The algorithm would
> not be able to consult a dictionary to do so.
>
> If you're interested in working on this, I'd be interested in hearing
> about it.
>
> mu'o mi'e la tsani

Nora also noticed that Pierre (phma) has his own effort at a
wordbreaking algorithm, "valfendi" on github. We are not sure where it
fits relative to brkwords put probably was later, though preceding
whatever was incorporated in camxes (one of the attempted post YACC
parsers which IIRC tried to handle speech streams)

lojbab

[Bob LeChevalier]

On 8/3/2020 4:04 PM, Jacob Thomas Errington wrote:

> I don't believe there is a proof out there, but it is by no means
> taboo to question whether Lojban actually achieves its stated goals.
>
> To prove that the morphology is unambiguous, one would have to write
> an algorithm that accepts an idealized representation of a speech
> stream and decomposes it into a sequence of words. The algorithm would
> not be able to consult a dictionary to do so.
>
> If you're interested in working on this, I'd be interested in hearing
> about it.

Nora reminds me that she wrote such a program as part of our efforts on
this. The program was called BRKWORDS and was written in Turbo-Pascal
3. It has some limitations on what it checks (permissible medials I
think is one area, and it may not do the slinku'i test properly), but it
does not consult a dictionary. It is up on the lojban.org website
somewhere (Google on "BRKWORDS" gives several hits with discussion of
the topic, and the same with .pas has the source code), or she can
supply her files to anyone who wishes to build on it.

lojbab