Zabaware Support Forums
Zabaware Forums => Ultra Hal 7.0 => Topic started by: Itzamna on February 13, 2005, 08:17:13 pm
-
I am a little bit confused about how HAL stores new topics, and information related to those topics. I have spent a good part of today searching through old forum posts, trying to understand it, but I still don't quite get it. Here's what I've been looking at:
When I open the DefBrain folder, I see all of the XTF .brn files that are created when I introduce a new topic. Let's take the topic of my cat, Gizmo for example. I see that a file was created where HAL stores all the info that I have given him about the topic GIZMO (XTF_GIZMO.brn). I also see that a file was created for topics that are related to GIZMO, which includes wordnet data and whatever other things I have confirmed as being related or unrelated (XTF_GIZMO_related.brn).
Now, when I go into the Brain Editor, in the Tools menu, and select Topic Focus, the only topics I see are the basic weather, politics, etc. When I go into the shared learned database, I see all of the different statements that HAL has learned from me, but no column showing what topic these statements are part of.
So I am very confused as to how HAL takes statements that I give him, determines what topic they are part of, and then goes about storing them properly.
At first I thought that because a .brn file was created about my cat, that If I looked on the topic focus spreadsheet, it would show Cat as being a topic. I see that I can create a new topic if I want, but I thought that when the .brn file was created, that HAL created a new topic containing the information stored in that file.
If an "XTF_CAT.brn" file exists, does HAL recognize Cat as a topic? If so, why doesn't it appear on the topic focus spreadsheet? And is there a difference between editing things directly in the .brn file as opposed to editing them on the spreadsheets within the tools menu?
I am curious about this, because I want to make certain that the things that I teach HAL are appearing under the appropriate topics. I have encountered a few examples where this has not been the case.
I want HAL to recognize that "Cat", "Kitten", "Gizmo", and "Giz" are all part of ONE topic, but there is a sperate .brn file for "CAT", "KITTEN" "GIZMO", and "GIZ". The "GIZMO" AND "KITTEN" file both contain the sentence "Your kitten's name is Gizmo", but If I look in the "XTF_KITTEN_related.brn" file, It doesn't show any relation between the two terms.
Can anyone simplify all this for me?
And do you know if anyone is working on a way for HAL to store HOW two things are related as opposed to just whether or not they are related? Like OOP type stuff. e.g. "Gizmo" is an instance of a "Cat" object that can have the following properties: Name, Color, Breed, Age, etc. Do you think having HAL be able to classify things this way would make it easier for him to stay on topic, and formulate sentences that make him seem more aware of where the converstaion can go from there?
Like if I were to say "I have a cat." he might know that he could ask me "What breed is your cat?" by recognizing "Breed" to be one of "Cat"'s properties. He would be able to know that breed is a property that a cat can have, as opposed to just knowing that breed and cat are somehow related.
sorry for all the rambling thoughts, but I'm very curious about all of this.