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

Don’t worry…Walter won’t care about any stinkin’ apostrophe. =)

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.

I see what you mean.  Without emotion,the best that chatbots can perform is (by analogy) 2D. However, with emotinal input then chatbots will become 3D.  NLP by itself is a weak substitute for ‘real’ human communications.

To Nathan’s point, I’ve pondered the notion that a chat bot, when not chatting, should be able to surf the web and ‘read’...and learn. 

Regards,
Chuck

Don’t worry…Walter won’t care about any stinkin’ apostrophe. =)

Then it really won’t completely understand what you mean.  Unless, are you going to try both possiblities and determine a likelyhood that the user just forgot an apostrope?  You can’t just assume, in all cases, that the apostrophe should have been there.

That’s what artificial intelligence is about

Exactly.  Generate the possiblities, then, based on context, determine the one that best fits. 

My point was basically, to be aware of that and code it in the bot.  That is, to be aware of the different meanings, with and without the apostrope, not to ignore the concept altogether.

I think so. That’s why I believe the main bottleneck of NLP will be the education problem in the coming future. At the early stage, we can not expect a chatbot to read normal articles on internet or in some corpus like SemCor when it has not enough general knowledge.

Like a human child, chatbot need a story book which fit for it’s level.

Chuck Bolin - Aug 20, 2010:

To Nathan’s point, I’ve pondered the notion that a chat bot, when not chatting, should be able to surf the web and ‘read’...and learn.

Nathan,

I can’t agree with you more!  I’m on the verge of completing all the core elements of my bot engine.  And it is going to have an enormous appetite for knowledge !    I was meaning to ask you, what sort of technology are you employing in your bot engine,  what sort of approach, neural networks, rule based, or perhaps your own design, or a combination?

Considering the name of the topic for this thread, has anyone watched the video

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

?

Any serious chat bot engine developer should give that a watch, it is quite good.  John Searle is quite a guy.

My chatbot engine is based on the approach of semantic network pattern recognition which is designed by myself. Actually, I believe it’s one of the right way to get to strong AI.

What’s the base approach of your chatbot? it’s nice to hear that you almost complete the core of the engine.

Victor Shulist - Aug 22, 2010:

Nathan,

I can’t agree with you more!  I’m on the verge of completing all the core elements of my bot engine.  And it is going to have an enormous appetite for knowledge !    I was meaning to ask you, what sort of technology are you employing in your bot engine,  what sort of approach, neural networks, rule based, or perhaps your own design, or a combination?

Considering the name of the topic for this thread, has anyone watched the video

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

?

Any serious chat bot engine developer should give that a watch, it is quite good.  John Searle is quite a guy.

That’s interesting.  My approach employs only my ideas as well.  I haven’t used any ideas that I have read about over the years in various AI books.

The only really ‘external’ information that I am pulling from is my handy dandy books on English grammar.  My approach, as you probably already guessed, is centered and firmly based on grammar.    The bot will begin its life knowing the entirety of English grammar, and yes, even bad grammar, so it is able to parse even though user uses the wrong constructs. 

For learning, I’m not going to have it ‘learn’ grammar, I will code that in so it has a ‘head start’, but for learning, I will teach it via NLP.  Think of it as an NLP ‘programming’ language, or I guess a better way to describe it, since it is not a convention procedure programing language, instead a better name would be ‘NLP knowledge acquiring language’.

It has been awhile since I updated this thread…

I just found this one… 2 years back, but it is interesting .. from Microsoft…

http://channel9.msdn.com/Blogs/Charles/Expert-to-Expert-Natural-Language-and-Computational-Linguistics

Reading back through this thread, I found interesting the examples of a pattern that humans would recognize instantly.

How could I get a computer to do the same? I have a crude “opposite agent”, which right now can do:

> what is the opposite of: That is a nice car.
That is not a nice car.

> what is the opposite of: I love this weather.
I do not love this weather.

> opposite: This apple is delicious.
This apple is not delicious.

So, with a little tweaking, I could get the computer to mimic the pattern, which is a first step towards recognition…

Next, I would associate the responses given in the examples with the generated opposites: “I hate this weather” = “I do not love this weather”, etc.

Then, I could create an agent to look for input-response pairs in the opposite relation. If several occur consecutively, the agent could be programmed to conclude the interlocutor is being contrarian!