[Data]

So you clever people, that I feel lucky to know  I have a simple question, from my simple mind, would appreciate a simple answer too 

IS there or are there any emerging technologies or programming that will better and/or replace pattern matching for chat bots or some new advanced Ai?

Or is the refinement of pattern matching the way to go?

I would much appreciate being brought up to date in a fashion that I can grasp 

Thanks.

[mendicott]

In addition to basic pattern matching in chatbots, there are also dialog systems based on a variety of statistical techniques such as Latent Semantic Indexing

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

as well as natural language systems based on Semantic Reasoning

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

[Data]

Thank-you for taking the time to answer my question mendicott, its helped clear the muddy water, so to speak.

So it seems only right to ask if there are any bots available today that use either of or part of these systems, does anybody know? 

[infurl]

There are no neatly packaged "bots" available to do any of these things but there is plenty of open source code available that you can try out. If you are serious then you should start learning how to program with CommonLisp as there is very little software of note in this field that was not built with some dialect of LISP. I'd recommend CLISP which is open source and very powerful, but if you must have a graphical IDE then you could try the free but limited personal edition of LispWorks instead.

Another good place to start looking would be OpenCyc and you might even be able to do some interesting things with it without knowing any LISP. Also, the famous SHRDLU program has been translated from the original MACLISP to a modern CommonLisp implementation. You can download it from http://www.semaphorecorp.com/misc/shrdlu.html although it only knows about coloured blocks and you might run out of things to talk about fairly quickly.

[mendicott]

There is one advanced agent available that I know of that might be fun to check out if you have good hardware available and time for the learning curve.  Its a sort of open source version of IBM Watson called The Ephyra Question Answering System, OpenEphyra at http://www.ephyra.info that retrieves accurate answers to natural language questions from the Web and other sources.

[Art]

I had emailed Terry Winograd many years ago to discuss his SHRDLU program and found both he and his program extremely interesting.

Considering the fact that he wrote and published the program around 1970 at M.I.T., which was oh...41 years ago +- !!  In all that time
we've really come no further in developing a system that could make inferences and logical interaction?!

We promote ourselves and pat our humanity for being so clever but are we really? There seems to be something holding us back from making
or taking that leap forward. So many of us that were so ingrained in the complexity back then are now ready for retirement or far removed
from the methodology and technology of today.

While there are some "bot" programs that will search the Internet for the "best" answer, the knowledge really isn't present. A clever fetcher
at best. Perhaps the day will come when our computers will be able to interact with humans on a nearly equal level. If it does come, how will
we react and will we really be ready for it?

How will we react when we discover that 'IT' is actually smarter than us? Hmmm....

[infurl]

The mention of ephyra reminded me of another knowledge base that you might find useful. It has the advantage that it is self-contained and can be downloaded in it's entirety from YAGO It has been compiled from an enormous list of verifiable facts extracted from Wikipedia via http://dbpedia.org/About and http://www.geonames.org/ It has also been aligned with WordNet for added value.

[Freddy]

That's a big download but interesting stuff.  I noted they say Watson (IBM) used it.

[ivanv]

hello!
did you know that automated theorem proving (http://en.wikipedia.org/wiki/Theorem_proving) can be used to find out a list of actions that would lead to wanted goal. really impressive, but the problem is to recognize such goal implicitly from input . yet it works fine with explicit queries.

[spydaz]

I think after reading so many papers on pattern matching ... I would say that  the Hearst algorithm... Is probably copied from others ... As she just formalized it by writing it and implementing it I a dissertation ...

All text extraction uses some form of pattern matching ...

I'm currently moving down the path of sentence identification and clause extraction ... In understanding different sentence types such as questions , declarations etc... Each have very different clause types yet the clause contains the information ... Pattern matching is useful when extracting components of the sentence yet still no true understanding of the text extracted is given as the data received is still general and contains no logic , plus you can't create every pattern in the world ... Which also means that if you want specific types of information you would need to build that pattern collection for each concept ...
Yet understanding h the clause enables you to get some logic from the sentence ... And identify if the sentence actually contains knowledge or seeks knowledge ... 

Sentences have specific clause patterns ... Each clause contains a sub pattern which contains the data . ... (Not to reveal to many secrets) pattern matching can identify the position. Of the indicator keywords ... Which can tell you if the clause contains the data you seek ...
If a then b if b then c
Given a conclusion = c

Actually doable in ultrahal ...

[ivan.moony]

I tihnk that NLP could be a direction to achieve smartbots.

Here is a parser, together with easy-to-read grammar rules (if you want to develop your own parser): EP4IR
Take a look at grammars and lexicon through links on the left side. The project is in public domain and it looks promissing to me.

P.S.
And here is another link which you could find usable if you are really into NLP: http://www.resourcebook.eu/searchll.php#. Just you have to search over google when you filter out the name of a NLP project.

[spydaz]

Interesting link ... Currently I'm searching for all the rules to determine a nounphrase and verbphrase  to build my own lexicon ...

Thank you ..
Dr moony !