[ILICH]

Hello everybody.
I'm trying to create my own chatbot using C#.
Now i'm on stage of implementation bot's understanding of the things that you saying to it.
My bot's mind is semantic network - bipartite graph of concepts and relations.
System can recognize part-of-speech of every word in sentence context, so i can build and recognize some higher level patterns like Subject-Predicate-Object and so on.

I just wondering what kind of semantic relations you've used in your own bots ?
How would you suggest to store following sentence in semantic network:
"My name is ILLIA"
Approach that i thought about was:
Suppose that i parsed all words and found/created associated concepts. So "my name" will correspond to logical "subject attrubute" and "subject" , "is" will correspond to predicate and "Illia" is proper noun, singular, and it is a logical "object". "my" word resolving depends on context, and if this is what user says - it means that it associated with user identity/personality concept.

My first intention to connect described concepts was to use HASA and ISA to resolve this.
So, we have 3 concepts: UserIdentity, Name, ILLIA
to reflect new information i should know that UserIdentity has Name property. Also i need to know that ILLIA is a name. 
So, ideally i need to have already 2 connections in robot's mind:
UserIdentity HASA Name, ILLIA ISA Name.
I should probably import such information from ontology or something .... (determine that ILLIA is singular proper noun).
Then i need to add new relation to wire up all those three Concepts ... and i dont know what is the best way to do this. Probably it depends on used inference model that will determine what property values some Concept has .... or ? i don't know. Did you faced with problem like that ? Do you have some thoughts how better to resolve it ?

So, probably my questions should look like this:
How you represent property relations in your semantic network ?
Do you distinguish intrinsic and external properties ?
How properties are related with HASA and ISA ? they should be related or not ?
 
Will appreciate any help about subj. Thanks.

Added: oh, sorry, just explored that UserIdentity ISA ILLIA is the right relation in my case because ILLIA is broader concept than UserIdentity

[Freddy]

Welcome to the site Ilich.

There are quite a few programmers here, hopefully someone will comment given time.

[ILICH]

Thanks Freddy ! Yea, maybe somebody already used similar ideas in their chatbot engines

[Bragi]

Perhaps this post on abstract and concrete knowledge might help.

[ILICH]

Hoho ! Thanks a lot. Very nice blog, very promising software ! 
Is it yours ? I should probably take a day to go through all things there

[Bragi]

Is it yours ?

yep, that's my little project.

[DaveMorton]

You may also want to check out the forum at chatbots.org, as well. The forums there are more geared to the technical/professional aspects of chatbot and AI design than the forums here are, so you may find some good information there, as well. I'm not "throwing stones" at these forums. Quite the contrary, I love the folks here like family. It's just that your post is a bit more technical in nature than the average post here.

Don't shoot me, Freddy. I'm just trying to help.

[Freddy]

*holsters gun*

Hehe no problem.  Yes we are more like enthusiasts here (not that programmers aren't) and end users although we do get a fair amount of concept discussion which is good.  I don't know how we could encourage this more, if there are any ideas feel free.  Over the past five years or so we have discussed many theoretical ideas about AI too.

But yes GeekCaveCreations is right, as with most things on the internet it is better to have more than one arrow for your bow.  I have spoken to Erwin a few times (coordinator of chatbots.org) and he is a really nice guy and is doing a great job over there, so I would say take a look there too.

By the way Ilich, something we are good at here is testing things.  There is nearly always someone who would take some time to try out the things people make and give feedback, so when you get that far let us know about it.

Best of luck  

P.S.  I'm glad you like it here Dave, it's great that you are here