Re: Natural Language
- Posted by "William J. Oney" <brown60 at ZIANET.COM> Mar 28, 1998
- 849 views
>Date: Sat, 28 Mar 1998 09:06:04 >To: mdeland at nwinfo.net >From: "William J. Oney" <brown60 at zianet.com> >Subject: Re: Natural Language > >Ok... this stuff is probably way over my head but... I love the subject :) >by the way, I'm new here, and my name is William Oney, and I'm 18... >and I basically screw around never finishing anything I program... oh well... hehe >anyhoo... > >>>teaching the human language to the computer with similar >>>expectations you would have of a learner of any foreign >>>language. >>hrmmmm...... >>firstly, when we say "human language" do we mean English? >>(petty point perhaps, but! IMHO English is the worst >>language to actually code a natural language in.... >>a better language is one that is more inherently "discrete" >>or "digital" as opposed to English, which is pretty wacked) > >---what about all of them? >---it's a computer right... don't have to worry about not remembering... or being too complicated... > >>secondly, teaching someone who *already* knows a language >>(any language) to be verbose/fluent in another language >>is *very* different ....*very different* indeed... from teaching >>someone who knows *no* language to speak any language >>at all. IMHO, the way we might want to think about >>implementing this is to use the context of teaching an >>infant to speak...since... in this context, the computer >>is much more akin to an infant than it is akin to an adult >>foreigner learning another language. > > > >>>I tried making a natural language system using English to >>>English interaction, but it didn't work very well for the very >>>reasons cited by other members of the list. >>and for perhaps the reason above? >> >>>But now I'm using >>>an idea originally proposed by the UN -- using an intermediate >>>language that logically defines grammar and vocabulary sorted >>>topically. >>much better... and perhaps more "digital" in nature. >>each word _meaning_ one thing and one thing only, >>each word _being_ one thing and one thing only. > >---so we're not talking about artificial intelligence? > >>no words that are nouns and verbs, or words like >>"read" and "read" (heh...which is which eh?) >>couldn't this intermediate language be *the* >>language itself? the actual natural language? >>and would this intermediate language be a >>true natural language, or as I said before, would it >>be in fact a "high-level" language instead? >> >>--snip-- >> >>here's another way this could be done...i think?... >>heh... >>using the premise of the infant, and perhaps >>some theory from neural network programming, >>supposeeeeeeeeeeeeeeeeeeeeeeeeeeee..... >>what if......BIG if.... >>what if we took a computer, gave it some sort of >>rudimentary interaction with its environment, >>let it begin making its own associations, letting >>it figure out from its surroundings what its own >>natural language will be, based on interaction >>with others using their own natural language. >>an infant learning to talk from its parents. >>much research has been done in this area, the >>biggest snafu so far that i have seen is >>processor power and patience... >>the first is now perhaps not much a problem >>anymore...the second...patience on the part >>of the researchers... no comment :) >>will this approach give us the result we desire? >>will it give us natural language programming? >>yes. >>will it give it to us rapidly? > >--- Here's the main reason I replied... > I'm probably being naive thinking not many people have thought of this, but > giving it the ability to browse the world wide web, not a very good teacher... > but more human-like language... eh... more rapid than feeding it by hand... > >>no. >>can this approach use euphoria? >>very much yes, and euphoria might just actually >>be the better of the programming language choices >>to do it in.... thoroughly variable data structures >>that expand and collapse dynamically at runtime >>and associate themselves in infinite potential >>connections between each other... >>hrmmmmmm.... >>sounds like neurons? >>sounds like how a baby learns? >>association 1 is experienced so, insert that >>association into a variable at such and such >>place without regard to what it might be, or how >>big (read, number of neurons needed to associate) >>that association might be... shove it in and euphoria >>just makes the sequence larger... handy for this... >> >>'nuff rambling ...'nuff food for thought :) >>Mike. >> > >well... I need to read a lot more before I can begin to offer much intelligence to your conversation, but it stimulates my brain... :) >