Victor, did you edit your post while I was creating mine? It somehow looks more… Um… enlarged, than before.

Thanks guys, I’m stuggling here with computers, making an interactive presentation (powerpoint plugins, virtual machines on Apple), on you’re attacking me . We’ll here we go:

Chuck Bolin - Aug 19, 2010:

I don’t quite agree. When a chat bot is developed that can chat or converse in a forum such as this…that will be quite a significant breakthrough.

Yes, that would be an enormous breakthrough. On my computer, I’ve a video interview with Hugh Loebner, who has been organizeing his Turing test for 20 years. He’s very sceptical about breakthroughs in this area. However, computers have passed the Turing test in specific areas, withs millions of areas to go.

I do believe speed in increasing intelligence has rise. I truely believe in the future of chatbots (that might be the reason why I started this whole thing ). However, I don’t believe this breakthrough will come from NLP.

Victor Shulist - Aug 19, 2010:

Oh, and I agree, you guys are great to discuss this stuff with !  I’m not always in 100% agreement, but that’s ok, as a friend of mine says… “healthy debates”     I am starting to learn what your definition of “pattern matching” is, and the more I do, the more I’m actually finding I do agree

Look forward to discuss with you in real life! We’ll organize conferences, and I personally it would be fun to a have a phililosophy night for non-philosophers (real philosophers assume you have knowledge of all philosphoy models, it’s impossible to talk with them)

From an academic perspective (we don’t have to use that here) pattern recognition is the art of recognizing partterns in information, however information is defined. It can be bits, it can be analogue waves. It can representation audio, video, sonar signals, telescopic waves from scape. What’t the logic in all this? Can we represent what will be next? Can we develop comppressing mechanisms? Can we develop security algorithms? It’s all about recognizing patterns.

When we talk about pattern recognition in the chatbot context, we often talk about matching words. Let me give another example of pattern recognition in academic sense:

Thats’ nice car!
I don’t like it…

This apple is delicious
I hate apples

I love this weather
The sun is too hot.

What’s the pattern? Humans see this right away and a dialogue will immediately change. That’s not the pattern recognition we’re talking about right now.

I believe that we’ll define emotional state, preferred emtional state, and input that feeds emotional state (such as warmth on our skin, sensory input on what we observe), that than NLP will be the next step.

So the research we’re currently developig is awesome, but the true breakthrough will come from another area and NLP will then accelarate the developments.

Dave Morton - Aug 19, 2010:

Now I may be wrong, but it’s my considered opinion that NLP IS pattern recognition, though it processes the recognized patterns in a far different manner than what “standard” pattern recognition does.

I don’t agree. NLP can be pattern recognition, but there are other approaches to NLP as well.

Pattern recognition is a concept from Information Theory, and is about recognizing patterns we’ve seen before, analogue or digital.

Researchers are now for example looking in patterns of chemicals and other substances in our blood in relation with deceases. Pattern recognition as well.

Exactly. when the basic NPL stuff is completed, we can extend the virtual world model to image to the real world.

Victor Shulist - Aug 19, 2010:

Similar to what Chuck is doing.

Also, when your core NLP stuff is completed, you can disconnect it from the virtual world and connect it to the real physical world via microphone, video camera, pressure sensors, GPS, etc.

Not to mention being able to freeze a virtual world and play events back.

Off topic: @Nathan, why do first type your reply and then the quote?

Erwin Van Lun - Aug 19, 2010:

We’ll organize conferences, and I personally it would be fun to a have a philosophy night for non-philosophers

You know something, I’m realizing that I am going to have to have a pattern matching algorithm as a ‘front end’ to my bot to handle:

a) misspelled words

b) words that are there but do not belong

c) missing words

for example, in your statement above…

        “I personally it would be…..”

I think you meant to have the word “think” between “personally” and “it”.

So that is case (c) above, but other times it could be cases (a) or (b).

Almost every email I receive has these issues.  And I have been realizing how I analyze these sentences.  My brain said ‘what the heck’ ... how did we get from “I personally” to “it would be…”  .. that doesn’t “sound right”.

How did my brain know that “I personally”  followed by “it would be”  didn’t make sense?

This is going to be especially challenging in bot engine design.  I think it will work the same basic way spell check does.  If it cannot parse the input string, it will have to consult some statistical database and see :

  ‘oh yes, “think” is very often followed by “I”.’,
  ok, then, let’s ‘pretend’  that “think” comes right after “I”, and run it through the bot again..

bingo ! ok, now we can parse it as “I personally think. . ..”

But all this will cost even MORE CPU time !  So yes, I can see ‘pattern matching’ aiding in NLP, it sure is not the alpha and omega of chatbot design, but it will help the core NLP processing

It is too bad that the only reason we would have to run through all these permutations of inserting new words the user left out, or removing extra words the user should not have entered, is only because of human error !!

Another example LOL….  chatbot would have to know to put “you” between “do” and “first”

Erwin Van Lun - Aug 19, 2010:

Off topic: @Nathan, why do first type your reply and then the quote?

As for Loebner prize, it has dropped in priority for my bot project considerably.  I find that it is actually at odds with creating a useful, intelligent chat bot, it would pull me off the proper course, and encourage me to ‘cut corners’ and attempt to pass with some glorified Eliza based bot.

In speech analysis they talk about prediction.

Do you like to *** in the mountains? Well, ski, hike, walk, what’s more?

To be honest, I belive the settings of Turing test have some problems. It can not reflect some important features of strong AI. And the winners always did not work based on the “true AI”. Turing test misleaded the research focus.

Before to make the computer chatting like a human, the most important issue is whether a computer can help the chatter to resolve some problems with natural language. A chatter may know he is talking with a computer, but it’s also valuable that the computer can help him to resolve problems in natural language.

Making the computer chatting like a human will be the last goal, not the first, not the most important. 

Erwin Van Lun - Aug 19, 2010:

On my computer, I’ve a video interview with Hugh Loebner, who has been organizeing his Turing test for 20 years. He’s very sceptical about breakthroughs in this area. However, computers have passed the Turing test in specific areas, withs millions of areas to go.

@Nathan, the Loebner prize has three levels:
-textual based
-voice based
-visual cues based (100.000 $ and end of the price)

Actually, you’re stating we need more levels.

What would those levels be?

Nathan . .. ABSOLUTELY !

The Loebner prize, as it is now, is actually impeding progress in AI and chatbot design by encouraging people to ‘cut corners’. 

It encourages people to develop ‘weak AI’ bots because it is simply not realistic to have all that functionality in a bot any time soon.

I have abandoned interest in that contest, I have found it is actually at odds with the path I want to go in order to develop my bot.

Nathan - that is the best statement I have ever read on this site !!  We need NLP working first, THEN we play the imitation game!

Also, why is the audio / visual test 100,000 and text based 25,000 ?

What the huge dollar amount difference?  We already have visual recognition software and audio recognition.  What does the form of input have to do with intelligence of the bot ?

What exactly IS the audio / visual Turing test anyway ?  Does audio / visual processing equate to ‘stronger’ AI ?

Animals have visual recognition, but their intellect is no where near humans

Hugh Loebner, skeptical ? No wonder, the conditions of the test are so unrealistic to pass, I don’t blame him, we haven’t, nor ever will probably in 50 years, see a winner.

It’s like having a car competition in the year 1900… i want air condition, DVD player, GPS,  power windows, etc….. to pass the test   Have a gradual Turing test, giving out 3,000$ for each significant improvement, not just to the best Eliza clone each year.

Even though I don’t think, even if he was correct about the Chinese Room, that a machine necessarily needs consciousness to be intelligent, consider the following link on ‘You Tube’ ..

Dr. John Searle :

http://www.youtube.com/watch?v=D10lAx3wfDk

Watch that video, and then Erwin, tell me that pattern matching is more important for AI than research than NLP !  

sorry, you won’t get that kind of functionality in a chatbot that Dr Searle talks about,  with pattern matching!!

What I told about is the textual based Turing test.
We need to set up a practical test for AI.

For example.

Read an article and answer some questions in natural language.

We can use different articles to test the understanding of AI, from story for children,  news to fictions.

We can also set different supervising class, such as unsupervised, limited supervised and unlimited supervised. The trainner may talk to the chatbot in natural language to explain and help the chatbot understand the articles.

Erwin Van Lun - Aug 20, 2010:

@Nathan, the Loebner prize has three levels:
-textual based
-voice based
-visual cues based (100.000 $ and end of the price)

Actually, you’re stating we need more levels.

What would those levels be?

What a GREAT idea Nathan!!!! Why did no one every come up with something like this?

It’s actually exactly how children learn language

Johns dad is called Peter
What is the name of Johns Dad?

Johns dad is celebrating his 60th anniversary tomorrow
What will happen to tomorrow to Johns dads (or the proper English sentence)

Actually we need a separate thread for this. Shall I split it? What should be the next title? leveled language test?

Erwin,  just F.Y.I.

it is   “John’s dad”  not “Johns dad”

John’s dad—means the dad of John

Johns dad—not meaningful, Johns means there are a number of Johns, 5 Johns, 10 Johns.

Now the weirdness of English,  “John’s dad” could also mean “John is dad”.  So this will be interesting for a chat bot to figure out if :

John’s dad = the dad of John

*OR*

is “John’s” a contraction (short form of)  “John is”.  Example: “John’s funny” does NOT mean “funny of John” , but rather “John is funny”

which does the user mean, based on context and meaning.

suppose a chatbot woud have corrected me
tx again