[Zero]

One Syntax to rule them all, One Syntax to write them,
One Syntax to bring them all and in the brightness express them

[Korrelan]

The spatiotemporal spike 

[kei10]

Hm? May I ask what is this about? 

[infurl]

https://en.wikipedia.org/wiki/Universal_grammar

[kei10]

Hang on, our emotion is instinctively a universal grammar. Except that it's different across different living things -- sort of. Over-smiling until your teeth exposes to a dog, and it will bite your face off, as to a dog, it thinks that it means seeking for a fight -- Google says. Emotion... is a universal grammar that is evolved as a survival mechanism. *Flies away*

[Korrelan]

I was thinking lower/ base level.

As binary logic is the 'universal grammar' modern computers utilize; the spatio-temporal spike is one of the base constituents of 'universal grammar' that intelligent biochemical machines use.

Even emotion is generated/ based on ye olde spikes.

Though I don't think this is what Zero had in mind



I would presume from the J. R. R. Tolkien reference that Zero has either discovered or designed a language syntax to encompass/ beat/ encapsulate/ express all others.

[infurl]

I would presume from the J. R. R. Tolkien reference that Zero has either discovered or designed a language syntax to encompass/ beat/ encapsulate/ express all others.

Eh? That's been done before. It's called the Chomsky Hierarchy.

https://en.wikipedia.org/wiki/Chomsky_hierarchy

regular expression < context free grammar < context sensitive grammar < recursively enumerable

Almost all natural languages including English can be defined using a context free grammar which is good, because context free grammars are computable in polynomial time, with the right algorithms. Context sensitive grammars take non-polynomial time to compute making them impractical for the most part.

The bulk of my AI work has been concerned with writing a GLR parser which is capable of parsing an unrestricted context free grammar. My parser can handle the tens of millions of grammar rules required for a language like English, and it is very fast, able to parse a book in the time it takes other parsers to process a sentence.

[Korrelan]

Hi Infurl

I had to Google GLR parser lol.

Your work/ project sounds very interesting. Is your system applicable to other kinds of formatted data streams; ocular information coming from the retina could be construed as a kind language with syntax… or is it alphanumerically based?

Once you have parsed the ambiguous text/ book; what is the resulting data format? How do you store the end results of the parsed book and what do you intend to do with the results? Is this for a ‘chatbot’, expert system; or is your ultimate goal an AGI?

 

[infurl]

Here's my most recent post on the subject which goes into any detail, but I've posted here and on chatbots.org quite a bit over the years.

http://aidreams.co.uk/forum/index.php?topic=10090.msg39064#msg39064

As you can see the output can be expressed as XML for inspection, debugging and processing, but calling the API directly as a library allows direct access to the parse tree for greater efficiency. Input is a stream of integers (i.e. unicode characters).

In addition to the extensive libraries that I've developed for managing large complex grammars, I've also developed a set of libraries for knowledge representation and problem solving using first order logic.

Most of my time is spent authoring the grammars and knowledge bases to be used by the software that I've written. My goal is to emulate and improve upon the best that is already out there.

[Korrelan]

Wow… you have certainly put some work in on this…

It would seem to me that your language/ knowledge parsing schema is geared mainly towards heuristic based knowledge systems.  Language parsing is currently well outside my sphere of knowledge/ experience though I can definitely see the potential uses; especially with the recent uptake by major businesses employing ‘chatbot’ sales/ help lines.

I'm still curious though; could/ can the parser be reversed? Can you foresee a way that the parsed/ acquired information could be re-formatted and verbally re-spoken by a TTS?

I envisage if you coded the extraction of information; you could also code a ‘reverse’ set of rules to drive an output stage?

 

[keghn]

i am interested in a  universal language connector. I envision it at a self organizing map, SOM, of a bag of words
of comparative features. And the would be the nouns.
 Then a real map mapping out the movement to and away from the nouns would form the verbs.

 But this only works if all the bot are of the same design.

[infurl]

Keghn have you considered using WordNet? It is available for many languages now and it has all the features that you describe.

https://wordnet.princeton.edu/

http://globalwordnet.org/wordnets-in-the-world/

[Art]

Well......if there is this "Universal Grammar" how come so many people can't use it?

[keghn]

 Much appreciated @Infuri.

  I am working on ground of symbols and this internal auto generated bot atom language. So i doing a bottom to top approach to language, speech, written letters, sign language, body language, and art.
 All this done in a unsupervised learning machine.

 I consider you a language expert and also her: 



 But no one knows hows primitive beginnings develops into complex communication. But i do. Or other do not care and
are happy dealing with just finished product.

[infurl]

keghn grounding is a difficult but crucial problem and there have been some interesting approaches to solving it. For my projects I'll be using ontologies such as SUMO and OpenCYC, but you might be interested in a particular project that has been running for several years now and which sounds a lot like yours.

It is called NELL for Never Ending Language Learning and it is a 'bot that is attempting to learn to read by scanning the world wide web. It is hosted at http://rtw.ml.cmu.edu/rtw/ and they have made all it's databases available for download by the curious.