http://computer.yourdictionary.com/production-system

production system - Technical definition of production system:

>>A computer system used to process an organization’s daily work. It implies a real-time operation and the most mission critical computer system in the enterprise. Contrast with a system used only for development and testing or for ad hoc inquiries and analysis.<<

Marcus Endicott - Dec 3, 2012:

Technical definition of production system:
A computer system used to process an organization’s daily work. It implies a real-time operation and the most mission critical computer system in the enterprise. Contrast with a system used only for development and testing or for ad hoc inquiries and analysis.

Wow, I’d never heard that definition of “production system” before. If that’s the definition you intended, then I largely misunderstood your post, therefore much of my response is rendered irrelevant. Sorry for the misunderstanding. (Darned ambiguous English language!)

P.S.—
I see that my Wikipedia link (which I thought I copied and pasted!) in my previous post was a nonexistent URL. Here is another attempt to post that URL that shows two different parse trees as output, depending on the user’s desired organizational preference:

http://en.wikipedia.org/wiki/Phrase_structure_grammar

Mark, the problem with your Wiki link was probably not you, but a “glitch” in the forum software that causes some URL’s with certain characters (most notably parentheses) to be improperly parsed. We’re aware of the problem, but have yet to find a solution. I’ll go back and “fix” the link in question, and post back here what I did to do so.

_________________________________________________________

Sadly, the problem isn’t subject to the “usual cheats” I’m often able to do to correct these sorts of problems. Many times, I can usually just replace the offending character with a “URL encoded” version, and that works fine. Apparently, this is not possible with parentheses, either left or right. I was able to “fix” the link, but only by having the link go to a URL on ~MY~ website, that re-directs to the proper place. This is ~NOT~ something that I would want to do on a regular basis, so I’ll be redoubling my efforts to figure out and fix the glitch in the forum.

Sorry for the inconvenience.

Andres Hohendahl - Dec 1, 2012:

I am building a more sophisticated patterns matching, based on a combination of common EBNF and some special operators, which are capable of targeting sintagmátic [syntactic?] segments,

Interesting. My initial impression is that BNF/EBNF is poorly suited to parsing natural language. For *representing* natural language they would be reasonable choices, but it seems to me that the goal of parsing is to determine the writer’s meaning as quickly as possible, and for that task the essence of what is needed is determination of the subject, object(s), and verb, as quickly as possible. I suppose BNF/EBNF would be useful for structuring the many different possible parsings, though, say in a list that would then be analyzed for correct interpretation. Interesting. That’s a representation I would not have considered for that application, but just might work.

http://en.wikipedia.org/wiki/Extended_Backus–Naur_Form
http://english.stackexchange.com/questions/32447/is-there-an-ebnf-that-covers-all-of-english

Andres Hohendahl - Dec 1, 2012:

coupled with good spell corrector phonetically enhanced

I’m pretty sure I know what you’re doing there. I once had a job where I sometimes looked up in a database toys that customers were seeking, and often experienced the frustration of trying to find a toy that a customer had called by an unknown spelling. One example was “high tech”: high tech? hi-tech? high-tek? hi-tek? Just one pronunciation generated at least 4 possibilities, not counting hyphenation, and the database didn’t recognize any of the spellings I tried. I suggested to the company that they hire me to program their system to handle such queries but evidently they decided I was better suited to doing such trivial work than doing skilled programming work. :-(

Andres Hohendahl - Dec 1, 2012:

they generate an AST TREE

An Abstract Syntax Tree (AST) is another structure that strikes me as more suitable for computer languages than natural language.

()
http://en.wikipedia.org/wiki/Abstract_syntax_tree
()
“For the trees used in computer science engineering, see Abstract syntax tree.”
http://en.wikipedia.org/wiki/Parse_tree

Andres Hohendahl - Dec 1, 2012:

For example it can evaluate a complex math expression, in the middle of a sentence, finding the exact position of the mostly logical math piece.

If you mean your system can distinguish between a math expression versus natural language and can encapsulate such embedded math expressions, I can see where that would be very useful when parsing technical documents.

Yes, your approach is definitely at the “sophisticated” end of parsing! Good luck, and let’s hope it pays off in better machine understanding! Thanks for letting us know that there are some serious parsing folks out there.

[P.S.—I changed some of your spellings to what I believed you meant.]

If [you] [mean] [your system] can [distinguish] between [a math expression] versus [natural language] and can [encapsulate] such [embedded math expressions], [I] can [see] where [that] would be [very useful] when [parsing] [technical documents].

is this any use? ...done more than just marking noun and verbs I suppose. Is that count as particular pattern that is calculatable as well in a sense?

If [you] [mean] [your system] can [distinguish] between [a math expression] versus [natural language] and can [encapsulate] such [embedded math expressions], [I] can [see] where [that] would be [very useful] when [parsing] [technical documents].

does this help ...done more than just marking noun and verbs I suppose. Is that count as particular pattern that is calculatable as well in a sense?

Sorry for duplicates! that was a error and your a time based thing… heheh!

Anyway just tossing a idea, maybe you see some thing here as I see ...things in at least blurry way, eh? Just don’t worry about my grammer it aint good I know. But my whole idea is not to show to the right grammaritical parshing but a trend here.

Only thing that your bot might be asking is “where it would be helpful?” even if it din’t know about the thing in first place. And I think a satisfactory answer can satisfy his urge easily as well. What do you guys think?

general arithematic operators that I see: if, can, betrween, versus, and, such

where, when I dont have to say anything ...smile

(Opinion request: Why do I think this auto correction feature of text control should be constitutionally banned ...smile)

NickyBlue (#2) - Feb 2, 2014:

Only thing that your bot might be asking is “where it would be helpful?” even if it din’t know about the thing in first place. And I think a satisfactory answer can satisfy his urge easily as well. What do you guys think?

I don’t understand what you mean here. What would the bot ask?

This topic was from before I joined the forum, but still relevant and interesting. I voted “I’m not making a chatbox per se” because chatting is really not my game, but I am developing a conversational AI with sophisticated parsing. Though I wouldn’t say that grammatical parsing makes for a good model of our language, my program does throw a hundred grammar rules at each word and its surroundings to determine each word’s type and role, before converting them to facts.
Whether that actually results in successful communication, is something we’ll yet have to see: Still plenty of miscommunications remain in the areas of ambiguity, shorthand and figurative speech. But that’s why it’s called development.

this last one before I go…

All I said is I just flagged nouns and verbs and you see the whole complex sentence fall into a sort of equation. You just have to enhance your operators domain here. And sentence itself is a rule. It don’t need any other than that. What ever rule you trying code is all based on this very rule that sentence itself denote in most cases. So is the parse tree. Sentence itself writern in that form or else it’ll be confusing for us as well.

may be now you can guess what I was driving at?

bye and good luck for your thing!

may be you are just thinking in term of single sentence. Just grab every sentence form any e-text or book. And then think what rules. I think you got all the rule there in it. isn’t it? You just have to flag overlapping that’s all. Its same as you trying code rule. Why don’t flag overlapps instead. And you have satisfactory working network I suppose.

bye

I’ve written rules to deal with both single sentences and combinations of sentences, whether separated by commas or not. There may basically be one noun-verb-noun language pattern, but that doesn’t translate to just one rule to program. Before you were able to understand and write English, you first had to learn many English grammar rules too, didn’t you?
Anyway, it is clear that main verbs and noun-phrases are key elements of language, I can only agree with that. I think I’m a bit ahead of your general idea there.

The other thing I wanted to comment with regards to the original post is that I don’t believe ten sentence patterns suffice, as they are literally the -basic- sentence patterns only.
Questions, commands, relative clauses, shorthand and Sherlock-Holmesian formulations contain even greater diversity in the order of elements. So I would only recommend them as a starting point, not as rules to stick to. I’ve found that grammar is merely a recommended framework for our language, but not quite the key to it.

I’ve found that grammar is merely a recommended framework for our language, but not quite the key to it.

To emulate Yoda:

When understanding this sentence a computer does, intelligent may it be, yes!

Ok, maybe not, but…

Exactly, Dave . Yoda dialect is a very good example.
Unfortunately my program repeated that as “Intelligent understands well. The sentence does a computer. Intelligent is intelligent.” Either it failed or it speaks Yoda language better than I understand it (nah, it failed). A challenge for later, perhaps.

Now it seems to me that all the AI would have to do is some POS tagging, then matching against simple structure rules, and if they fail, arrange the tagged input to match known accepted patterns and re-test, yes? Wouldn’t that possibly produce an “understandable” input?