Re: Announcing syntreenet: A Python library to build scalable production rule systems

>
> Hello Enrique,

Thanks for sharing!

Le sam. 6 juil. 2019 à 20:53, Enrique Pérez Arnaud <enrique@cazalla.net> a
écrit :

> Hi all,
>
> Perhaps some of you might be interested in this library I've released, that
> might be used to build production rule systems, with the peculiarity that
> the cost of matching a fact to the knowledge base would be logarithmic in
> the size of said knowledge base, measured as the number of rules plus the
> number of facts in working memory. As far as I know, the current state of
> the art is polynomial.
>
> https://gitlab.com/enriquepablo/syntreenet
>
>
I am interested in this work. Mind the fact that I am a newbie and I might
ask trivial questions.

First, I can not tell the importance of this work compared to existing
solutions
even if what is described in the README seems promising.

Questions:

A) What about the name syntreenet?

B) What are Syntagm?

C) I read the following:

"syntreenet knows nothing about syntax. Facts are a black box with whatever
> internal structure that is appropriate for their universe of discourse.
> syntreenet never composes a fact."
>

What does syntax means in this context?

It seems to me for the purpose of the demonstration, it would be easier to
avoid to be generic and stick to strings for facts and rules and have a
convention for variables. What do you think?

Based on this log:

adding rule "X1 isa X2; X2 is X3 -> X1 isa X3"adding rule "X1 is X2;
X2 is X3 -> X1 is X3"adding fact "animal is thing"adding rule "X1 isa
animal -> X1 isa thing"adding rule "thing is X3 -> animal is
X3"*adding rule "X1 is animal -> X1 is thing"*

It seems like the algorithm rely on pre-computing rules based on existing
rules and facts, isn't it?
It seems to me the inference process starts before the kb is actually asked
something.


Thanks!
>
> --
> Enrique Pérez Arnaud
>


-- 
Amirouche ~ amz3 ~ https://hyper.dev