Talk:Inference engine
![]() | Business Start‑class Mid‑importance | |||||||||
|
Comment
This article ends with:
"the inference engine model allows a more complete separation of the knowledge (in the rules) from the control (the inference engine)."
What about separation from the data (view)? Isn't that just as important?
- I agree, it was too wp:POV. I've redone the page and taken that out. RedDog (talk) 19:32, 6 December 2013 (UTC)
Don't Merge
An inference engine may infer facts from existing facts/non-fact inputs. Business Rules Engines and Production Engines are based on the use of conditional rules. An inference engine may be constructed using rules. Howeve, a inference engine may be constructed using Neural Net and other technologies for pattern recognition to infer facts, and in such cases can include adbuctive logic. In which case it would clearly not be a Business Rules Engine nor would it be a Production Engine. In that respect, Business Rules Engine or Production Engine can be considered to be a sub catagory of an Inference Engine. 192.91.172.36 (talk) 20:17, 9 June 2009 (UTC)
Agree. Don`t merge, because "Inference engine" is 100% topic about programing techniques which define logic programming, automated reasoning, belief networks (aka Bayesian networks) and etc, etc. So "Inference engine" is strictly an IT topic, and subtopic of "artificial intelligence". In contrast "Business Rules Engines" is very narrow subject and may or may not be related to specific inference engine. —Preceding unsigned comment added by 212.59.0.201 (talk) 13:49, 20 August 2009 (UTC)
Merge
Production system, Business rules engine, production system, inference engine, all those are basically the same notion, alll these articles needs to be merged--Kompere (talk) 14:56, 20 December 2007 (UTC)
- I do not think that an inference engine is actually a sub-type of reasoning engine – the terms are used interchangeably in many contexts! For instance, I got [1] and [2] from the first page of Google hits. Hence, inference engine should not redirect to a production-rule topic, but rather to reasoning engine (which should then clarify the different applications fields early on). But inference engine's current production-rule-heavy content could indeed be merged into the suggested production rule articles. --Markus Krötzsch 21:49, 16 January 2008 (UTC)
Can the articles be clearly contrasted then, please? I.e. have paragraphs which explicitly say why this is different from the other, maybe with some examples. Especially: what is done in reasoning (inferencing?) engines that can't be done with a rule system? --Shepard (talk) 15:35, 12 May 2009 (UTC)
- I can not in any way agree with the statement that "Production system, Business rules engine, production system, inference engine, all those are basically the same notion". Sorry my friend, there are basic computer science errors in that statement. A production system is almost always forward chaining while inference may be Backward chaining as well, or opportunistic reasoning. Production rules are most often exact, while backwards rules may often be inexact, etc. Some of these are distinct concepts, and need separate articles at some point. None of the articles is well written by the way.... History2007 (talk) 00:58, 24 February 2010 (UTC)
!Merge
I do not second the merging of this article with business rules engine. Business rules engines are often coupled with user interfaces and serve to provide high-level tooling to automate decision making in business flows. Underneath the covers, an inference engine is often used. An inference engine can be used for purposes outside that of business rules. In the case of Semantic Web technologies, an inference engine is used to complete the missing but inferable data during a SPARQL query. —Preceding unsigned comment added by 86.31.55.154 (talk) 18:42, 16 October 2008 (UTC)
Do not merge, please
I do not agree with merging of this article. Inference engine is an important part of the semantic web. The article should be expanded a bit to clarify the differences perhaps, but merging it would be a mistake. It's not exactly the same thing as business rules or production systems.
Dragoniel Silverwing (talk) 07:16, 3 May 2009 (UTC)
- Please see my comment above. The merge suggestion is technically incorrect. History2007 (talk) 01:07, 24 February 2010 (UTC)
Quality
This article is very low quality, often incorrect and can not be relied upon. Needs a rewrite some day if I get to it. History2007 (talk) 01:12, 24 February 2010 (UTC)
- I agree. I rewrote the introduction and gave it a decent reference. What is currently there still needs major work or perhaps just a complete rewrite. The small section on "data driven vs procedural" I think should just be taken out. There is no reference (there were no references at all when I started editing, there wasn't even code to create a reflist I had to add it). I see what the person is getting at with the "data driven" section but I have two major problems with it. First, the writing is just awful, barely comprehensible. Second, it reads like an argument that rules are in some metaphysical sense "better" than OO or procedural code. I remember people having such silly arguments in the 80's but no decent software engineer takes those arguments seriously. Rules are good for some things, procedural, and objects for others. To claim that there is something abstractly "better" about rules is just wrong. From a formal standpoint it's all a turing machine and all equivalently powerful. It's true that an inference engine can eliminate or reduce some kinds of errors but it can also increase the probability of other kinds. In these articles we should never come down as X is better than Y but explain when and why one formalism or technology is more or less appropriate to certain kinds of problems. For example an Inference Engine for problems with complex logic that is volatile and mission critical. Sorry, that was a bit of a rant. Anyway, I'm planning on rewriting or just removing most of what is left from the original article and adding more references. RedDog (talk) 13:20, 5 December 2013 (UTC)
- I've re-written the article now. I kept some of the ideas but rewrote them in a way I thought made sense, took what what didn't, and added several references. RedDog (talk) 19:33, 6 December 2013 (UTC)
Do Not Merge
An Inference Engine is not the same as a production system nor a business rules engine. All three are different things. A production system is a kind of inference engine. It's the most popular kind but not the only kind. And a business rules engine is usually a production system but not always. Business rules can be expressed as process diagrams as well. And special kinds of inference engines called classifiers work on the subsumption relations in the data rather than on rules. Anyway, an inference engine definitely should be it's own thing. RedDog (talk) 04:54, 5 December 2013 (UTC)
major problems with current article
my keyboard is screwed up so i have to stick to small letters. since i can't use caps i can't rewrite this right now but i thought i would at least document some major problems i see with the current article in case anyone wants to discuss it before i change it. the whole section on data driven i think pretty much needs to go, as i wrote elsewhere it's just a religious type of argument that "rules are better than procedural" which is of no value and there are no refs. similarly the architecture section i think is also flawed. i kind of understand what it's saying but i don't recall ever seeing the architecture of an inference engine described that way and again no references. also in the big section on the recognize act cycle when it starts out talking about FSA i think that's wrong. i'm not an expert on FSA but that doesn't sound like a state machine as described. "match rules" is the description of a process not a state, it would make sense for a transition but not a state description, in any case bringing in FSA's is not needed and just something to make it sound formal. finally, the whole recognize act section is really about forward chaining only which is only one half of most inference engines. i'm tempted to just remove everything and just leave the intro text which is basic but at least sourced and reasonable but i want to give it some more thought and fix my keyboard first and see if anyone else has opinions RedDog (talk) 21:38, 5 December 2013 (UTC)
- I've rewritten the whole article. I saved a bit of what was there but wrote it in a form I thought made sense and was backed up by references RedDog (talk) 18:06, 6 December 2013 (UTC)