Inference systems?

[dtenny]

Seems like the list should include a section on inference tools (forward and backward chaining), I didn't see them when I looked today.

Some links I have examined (not necessarily recently) and thought were interesting, but that I have not used and so cannot vouch for, include:

• https://lisa.sourceforge.net/
• https://github.com/Incanus3/ExiL
• https://github.com/sharplispers/ops5

Presumably there are other Rete net implementations too. Which ones are good?

The links below I know little about and just quicksearched today to try to flesh out the topic:

• cl-prolog2 (in quicklisp) (interface to external prologs)
• https://github.com/guitarvydas/wam (Warren Abstract Machine, but not Prolog, which Prolog uses this implementation?)
• ? is there a native optimized ISO prolog in CL ?

Perhaps someone with experience using these or similar things can add some credible recommendations. Perhaps these things will become trendy once again in the age of LLM's.

[fstamour]

Related: https://gitlab.common-lisp.net/cl-unification/cl-unification

[surabax]

There is https://github.com/heshrobe/joshua-dist that can do both forward and backward chaining, but the license is not clear.

[guitarvydas]

WAM is Prolog. WAM is the defacto standard for a Prolog VM. My cl-wam repo was my attempt at understanding Ait-Kaci's tutorial about WAM and seeing if I could build my own in Common Lisp. I was testing it incrementally in bits and pieces and think that I veered off onto some other project just before I finished. Apparently, I once understood the WAM. I would be happy to kibbitz with anyone who would like to continue finishing this thing.

AFAIK, WAM is used in gnu prolog. I think that gnu prolog creates a WAM in Prolog. I think that gnu prolog is open source. I didn't go deeply in that direction.

IMO, Prolog is more interesting as a DSL for pattern-matching. I've looked at the main implementations of Prolog in Common Lisp and found that Nils Holm's version in Scheme is the most straight-forward. I've ported it to CL (and to JS). https://github.com/guitarvydas/cl-holm-prolog (the repo probably needs to be restructured to be more instantly usable - again I would be glad to kibbitz with anyone interested).

[vindarel]

related, an excerpt from https://european-lisp-symposium.org/static/proceedings/2024.pdf "Grants4Companies" paper:

---

While there are Prolog libraries for
Common Lisp16, and even one that allows converting Lisp data to
Prolog syntax and forward the result to an implementation17, none
of them completely fulfilled the requirements:
• bidirectional communication,
• parsing the Prolog output to provide a highlighted/clickable
display in the web UI,
• multiple parallel, independent sessions to concurrently test
different analyses,
• ability to export the Prolog input data for use in a separated
(offline), ISO-conformant Prolog system.

So we ended up with our own implementation, using Scryer Prolog18
as backend.

16See, e.g., [2 ], https://www.lispworks.com/documentation/lw445/KW-W/html/kwprolog-w-152.htm, https://github.com/nikodemus/screamer
17https://github.com/cl-model-languages/cl-prolog2

---

and they don't give a link to their project.

Paper co-authored by Philip Marek (https://github.com/phmarek).

[dtenny]

Trying to figure out if PowerLoom is used these days. There doesn't seem to be a page anywhere talking about who is using it, but it seems like it's being maintained. If anybody has knowledge on this. It's mentioned on cliki.net too.

[dtenny]

There is https://github.com/heshrobe/joshua-dist that can do both forward and backward chaining, but the license is not clear.

A search for additional information on Joshua turned up almost nothing, except for this link on reddit (with a video I have yet to watch), and in its comments a link to 2005 OpenCourseWare materials which includes pointers to Joshua reference and user guide PDF's. https://www.reddit.com/r/lisp/comments/704clr/symbolics_joshua_expert_system_operation_with/

https://ocw.mit.edu/courses/6-871-knowledge-based-applications-systems-spring-2005/pages/assignments/

It's interesting, but I'm thinking this isn't exactly mainstream inference tooling to link in awesome-cl.