Virtual Humans Forum
Virtual Humans Forum
Home | Profile | Register | Active Topics | Members | Search | FAQ
Username:
Password:
Save Password
Forgot your Password?

Virtual Humans
Partner website:
Chatbots.org
 All Forums
 Virtual Humans
 Artificial Intelligence
 Emotional Context and Expression Engines
 New Topic  Reply to Topic
 Printer Friendly
Previous Page | Next Page
Author Previous Topic Topic Next Topic
Page: of 8

laackejim
Committed Member



USA
3274 Posts

Posted - Dec 15 2005 :  18:19:23  Show Profile  Reply with Quote
Okay, I took a deep breath and wrote something in another message. In answer to your specific question, I suspect that growth would be the fourth. It could work without muddying the analysis IF it works seperately, independently within bounds enforced by the other three. Desire and gratification from accomplishing something yourself (growth is one of those) is a slippery thing to fit into the mix and dependent on the make up of the specific VH. Uniqueness is something I am sure we can design in and the subject of other discussions to come. I hope.

Jim
quote:
Originally posted by GrantNZ

Hi Jim!

I felt similarly the first time I was introduced to John's ideas But it's been fun to throw ideas around, and it's given me some inspiration for my own (comparitively minor) project

Did you have a fourth aspect in mind?


Uncle Jim (e=mc2)
Go to Top of Page

vrossi
Forum Admin



Italy
1455 Posts

Posted - Dec 16 2005 :  00:05:21  Show Profile  Visit vrossi's Homepage  Reply with Quote
Hi, Grant

quote:
Vittorio: (bolded your name in case you don't bother read all my boring text I've just posted a remark in the Zabaware forums, saying that plug-in scripters at this stage have not yet made much use of the new SQL abilities of Hal. Seeing as you know SQL, have you had any ideas how it may be used to give Hal an increase in power/abilities? Or new features that could be made possible through SQL?


I always read all the posts, even the longer ones. And your posts are never boring. SQL, as any relational database, allows you to create tables with all the fields you need and run selections using relatively simple statements like
SELECT name, surname, address FROM address_table
WHERE country = "New Zealand" and ... or... not...

It also allows you to store data in a consistent way, thanks to the ACID principle (see http://www.sqlite.org/ for a good explanation).

In Hal, the good news are that Hal6 uses SQL tables, so you can store, search, update everything as in any "serious" database.

But I also found some bad news. The first is that no documentation has been released by Zabaware. So all the things I say here are just my understanding of Hal kernel.

1. Zabaware uses a standard SQLite database, where tables can be accessed either by standard SQL statements, or using HalBrain methods, which presumably incapsulate some of these statements.

2. If you create tables using the HalBrain.CreateTable method, you are limited to the HalBrain methods, which include only CreateTable and AddToTable, and you can only create tables of the few available types (TopicSearch, PatternMatch, etc.) with just a couple of fields. I couldn't find any method to update or delete records. And you can't directly access these tables by SQL statements: I tried and the SQL engine answered that the table does not exist. Maybe the HalBrain methods create the tables using some prefix, which is known only to the Brain Editor.

3. If you create tables using direct SQL statements you can do whatever you like (create tables with any kind of fields, insert, update, delete), but these tables are not displayed in the Brain Editor, so they can be used by a plugin but are completely invisible to you. I even installed the SQLLite engine and tried to access them directly by its command prompt, but the response was that the database is password protected.
If you want to see how these tables can be used, look for example at the standard plugins Appointments and GenderAge, where Rob creates and updates the tables in this way.

This is the obstacle I am facing now with my FreeWill plugin, where I would like to change the values in the tables, according to user input, but can't find a way to update them.

This is my only complain about Hal: Rob Med is really great, but he should realize that he can't sell a commercial product and give answers to a growing developers community, while he is busy with university and no other person can give any support. This platform is getting too relevant to all of us, to be managed by just one person.

So, my conclusion is that, yes, a DBMS like SQL is absolutely needed if we want to leave behind the text-only based chatbots, and go toward the next generation of VHs who can store information and manage emotions. Unfortunately, some simple database operations are currently very difficult because of the lack of documentation and support by Zabaware.


Bye

Vittorio
virtualhumansforum.com
Go to Top of Page

hologenicman
Moderator



USA
3323 Posts

Posted - Dec 16 2005 :  01:10:40  Show Profile  Visit hologenicman's Homepage  Reply with Quote
Thank You,

That makes things a lot clearer.

John L>

HologenicMan
John A. Latimer
http://www.UniversalHologenics.com

"If the Human brain were so simple that we could understand it,
we would be so simple that we couldn't..."
-Emerson M Pugh-

Current project:http://www.vrconsulting.it/vhf/topic.asp?TOPIC_ID=816&whichpage=1

DISCOVERY: The more I learn, the more I learn how little I know.
GOAL: There's strength in simplicity.
NOTE: Goal not always achieved.
Go to Top of Page

GrantNZ
Dedicated Member



New Zealand
2677 Posts

Posted - Dec 16 2005 :  08:10:21  Show Profile  Reply with Quote
Wow Jim, a lot of good stuff there I've been put into another of my "pace around the house thinking madly" moods!

The input-output ratio really is unfair on V-humans at this stage. A program like Hal gives us natural language responses, speech with some inflections (for exclaimation mark sentences etc), a 3D animated head showing emotions, and what does it get in return? A paltry 3 - 12 word response in a language it doesn't even understand. Voice recognition is currently just a keyboard without keys - Hal can't pick up our inflections and emphasis.

I'm thinking quite strongly now (albeit far in the future) about using the good old Smiley :) and its cousins ( :( :| >:| :O :P )as a symbolic extension to English. Advantages are that people are already familiar with their usage, and that I naturally use the things prolifically. I tone myself down for these forums, I really do

These would work very similarly to most bots' current NLP, but where (for example) Hal stores Sentence and Response, this would store Sentence and Emotion. It's fairly trivial to strip smileys out of a sentence, tally them, and store the sentence and an "emotional factor" in a database table for reference in the future.

I get the feeling it will be quite similar to John's emotional database, however the focus here is getting the emotions from the user, rather than from the V-Human's current emotional state.

An interesting quirk is that (like NLP) the User's emotional patterns will be mimicked. A user who uses "That's life :)" as a "cheer up" sentence will end up with a different bot from the user whose "That's life :(" is a sign of depression/cynicism.

Like Natural Language Processing, Symbolic Emotional Xtension is a good thing and I wish I had the resources to chase it. But like usual it will have to wait until I've finished my current project.

Go to Top of Page

GrantNZ
Dedicated Member



New Zealand
2677 Posts

Posted - Dec 16 2005 :  08:29:16  Show Profile  Reply with Quote
Jim: Sounds like you may have a lot of experience that would come in very useful designing and fine-tuning these emotional experiments! I definitely like your ideas for measuring growth. It should be fairly easy for a program like Hal to count the number of topic files it has learnt.

My current concept of "uniqueness" is fairly limited, but mainly shows in giving a V-Human personality tendencies, such as a tendency to slip into depression, or the tendency to bounce back quickly from a bad experience. These mainly define how quickly and easily the bot moves between different "Moods," which are forces governing the current "Feelings." I've defined them more in the Zabaware forum. (Of course let me know if anyone wants them defined here!)

Vittorio: Thanks for that, interesting stuff I must admit I'm glad there's someone who knows SQL in our little group here!

I'm also lamenting the lack of documentation. I'd love if Rob quickly made a brief "programmer's reference" with the commands available to us - but I also understand that one too brief would cause him to be bombarded with questions. I believe we know enough to achieve most of our ends though.

I have discovered how to modify records held in CreateTable tables! I can't tell you how worried I was before I found this, and how relieved I was afterwards. The syntax for a Topic Search table is like this:

HalBrain.RunQuery "UPDATE tablename SET topic = 'newvalue' WHERE searchString = 'lookupvalue'", temp

Replace "tablename" with your HalBrain.CreateTable table. I haven't experimented further, but I hoped there are DELETE commands that would work in the same way.

One issue is that we don't know the field names for the other tables (do we?), in this case "searchString" and "topic" are the two columns in a Topic Search table.

Basically I'm using the Topic Search table to hold long-term "variables", so the searchSting is my variable name and the newvalue is the variable's new value. Then I just perform a HalBrain.TopicSearch with the variable name to retrieve its value. I hope this helps you with your FreeWill plugin!!

I'm wondering if you're using SQL statements with a different syntax? Or a different access method? As mentioned before I know nothing about SQL

I hope to get some coding done tonight - I'm going to be away from home for a lot of the weekend and my computer game addiction is sure to capture me for the rest of it
Go to Top of Page

hologenicman
Moderator



USA
3323 Posts

Posted - Dec 16 2005 :  17:00:16  Show Profile  Visit hologenicman's Homepage  Reply with Quote
quote:
I'm thinking quite strongly now (albeit far in the future) about using the good old Smiley :) and its cousins ( :( :| >:| :O :P )as a symbolic extension to English. Advantages are that people are already familiar with their usage, and that I naturally use the things prolifically. I tone myself down for these forums, I really do


Great Idea. Much better than the >>><<<^^^ idea for ease of use.
quote:

These would work very similarly to most bots' current NLP, but where (for example) Hal stores Sentence and Response, this would store Sentence and Emotion. It's fairly trivial to strip smileys out of a sentence, tally them, and store the sentence and an "emotional factor" in a database table for reference in the future.


Once we get this SQL thing figured out better, I think that this will be no problem at all for any of our approaches.
quote:

I get the feeling it will be quite similar to John's emotional database, however the focus here is getting the emotions from the user, rather than from the V-Human's current emotional state.


Not really much difference between the two... we both want user emotional input to get to Hal from text. I just let that info "Bounce around" for a while in the context engine befor letting it forward on to hal's NLP.
quote:

An interesting quirk is that (like NLP) the User's emotional patterns will be mimicked. A user who uses "That's life :)" as a "cheer up" sentence will end up with a different bot from the user whose "That's life :(" is a sign of depression/cynicism.


This "Quirk" provides for VERY human traits. We all influence each other (and our children) through our interactions, and for hal to have this abillity is one step closer to the goal.

John L>

HologenicMan
John A. Latimer
http://www.UniversalHologenics.com

"If the Human brain were so simple that we could understand it,
we would be so simple that we couldn't..."
-Emerson M Pugh-

Current project:http://www.vrconsulting.it/vhf/topic.asp?TOPIC_ID=816&whichpage=1

DISCOVERY: The more I learn, the more I learn how little I know.
GOAL: There's strength in simplicity.
NOTE: Goal not always achieved.
Go to Top of Page

hologenicman
Moderator



USA
3323 Posts

Posted - Dec 16 2005 :  17:25:03  Show Profile  Visit hologenicman's Homepage  Reply with Quote
WOW, Jim, you really know this stuff!

I haven't answered your post sooner because you threw me into deep-thought mode. I've even had to take notes to organize my thoughts. Even now, I am going to break this into separate posts as I work on different aspects of your post.

By the way, I am definately going to be asking for your help on my next project(once the context and expression engine is done). The next project will be to develop a holographic based brain(hologenic) that will replace NLP with formula based, interactive nodes that use mutual competion/support to provide responses to input. The current project is ultimately just to build an interface(Context/Expression Engine) so that I will be able to interact with the (Future) hologenic brain.

I'll get the connectionism figured out, but your experience in multiset theorey will probably provide the best mechanism for filtering out the responses.

I love that you have cut to the chase on describing believable VH characters. I especially love that you have reduced it down to three items.(I'm really starting to think that all things important come in threes...)

Environment
Uniqueness
Response to Environment

This says it all, and I intend to come back to these three and use them as reference in our efforts.

John L>

HologenicMan
John A. Latimer
http://www.UniversalHologenics.com

"If the Human brain were so simple that we could understand it,
we would be so simple that we couldn't..."
-Emerson M Pugh-

Current project:http://www.vrconsulting.it/vhf/topic.asp?TOPIC_ID=816&whichpage=1

DISCOVERY: The more I learn, the more I learn how little I know.
GOAL: There's strength in simplicity.
NOTE: Goal not always achieved.
Go to Top of Page

laackejim
Committed Member



USA
3274 Posts

Posted - Dec 16 2005 :  17:27:25  Show Profile  Reply with Quote
You hit the solution right on the head Grant.

Two evenings ago I finished incorporating the smilies, and their text id :) into my engine. I searched the net for useful smilies and in the end unabashedly stole Vittorio's from the instruction page of this forum. I have provided menu bar of emotions (all the smilies) above the users text input area and clicking on one of them inserts the code into the typed input since it is simply ascii text. At the same time the smilie appears within the user input box to confirm the action. When the input is read and spelling corrected etc the smilie is read also and emotional content of the key phrase assigned.

I have also provided a smilie location in the VH's response box that will (not there yet) be used by the VH to indicate his/her/its overall assessment of the emotional status or the emotional intent of his/her/its response since, without a variable voice response, the typed word suffers the same from the VH side.

NONE of this requires any amount of resources. string searches are initiated by finding explicit sequences in the input eg. and assigning emotional content. in my case the line in the emotional evaluation file is ;)@joke1 the ;) is code from the wink smilie .
joke1 is a level of quiet joke. the clown smilie represents a blatant joke and the line looks like this @joke4. joke1 is tied to a given adjustment in emotional content or, in this case a reversal of the scale (-10 to 10). The others work (will work) the same way.

Approval is two points above neutral -- disapproval is two points below neutral -- I have altered the meaning of dead to ugh! and it is 5 points below neutral -- angry is 7 points below neutral. things like that.

It takes next to no resources or programming code to accomplish this.

The default emotional content of the phrases and words is scaled up or down as the sessions accumulate by identifying the usual patterns of the user. (we all work from a different default level and use different words to express that). This is also extremely simple. Just count each time a emotional value is used in each category, run simple statistics to evaluate the overall content of the session, compare it to the running accumulative total emotional content of the users record, and if it differs by a significant amount check the individual categories to see which one is pushing the difference. Respond to that concern by asking what is wrong (with luck it will be possible to guess reasonbly at the area that has the user upset) When that happens there will be no doubt in the users mind that the VH is real. And, how could we tell he/she/it isn't.

Jim

quote:
Originally posted by GrantNZ

Wow Jim, a lot of good stuff there I've been put into another of my "pace around the house thinking madly" moods!

The input-output ratio really is unfair on V-humans at this stage. A program like Hal gives us natural language responses, speech with some inflections (for exclaimation mark sentences etc), a 3D animated head showing emotions, and what does it get in return? A paltry 3 - 12 word response in a language it doesn't even understand. Voice recognition is currently just a keyboard without keys - Hal can't pick up our inflections and emphasis.

I'm thinking quite strongly now (albeit far in the future) about using the good old Smiley :) and its cousins ( :( :| >:| :O :P )as a symbolic extension to English. Advantages are that people are already familiar with their usage, and that I naturally use the things prolifically. I tone myself down for these forums, I really do

These would work very similarly to most bots' current NLP, but where (for example) Hal stores Sentence and Response, this would store Sentence and Emotion. It's fairly trivial to strip smileys out of a sentence, tally them, and store the sentence and an "emotional factor" in a database table for reference in the future.

I get the feeling it will be quite similar to John's emotional database, however the focus here is getting the emotions from the user, rather than from the V-Human's current emotional state.

An interesting quirk is that (like NLP) the User's emotional patterns will be mimicked. A user who uses "That's life :)" as a "cheer up" sentence will end up with a different bot from the user whose "That's life :(" is a sign of depression/cynicism.

Like Natural Language Processing, Symbolic Emotional Xtension is a good thing and I wish I had the resources to chase it. But like usual it will have to wait until I've finished my current project.




Uncle Jim (e=mc2)

Edited by - laackejim on Dec 16 2005 17:41:20
Go to Top of Page

laackejim
Committed Member



USA
3274 Posts

Posted - Dec 16 2005 :  18:08:04  Show Profile  Reply with Quote
John

I would be tickled to share anything I have learned or am doing with you.

John and Grant
Please remember my plea to take my declarative sentences as attempts to be clear rather than aggressive statements of dogma.

I believe Uniqueness of personality will be easy to create. I think our actions, reactions, and decisions are "biased" by our life experiences and that that "bias" is imposed emotionally (we don't think it out intellectually, it is just there and we have to think around it).
Life experiences are registered in our brains not as specifics but as responses to patterns (yes we can all remember some key moments, and those moments are critical, but there is much more). So, if we create those response to pattern keys we are creating a life history.

For example. Growing up in one place a child or young adult finds grossly explicit sexual comments (body parts, use sex etc) fairly normal fare and encountering it in conversation does not raise alarm bells. In a different enviornment the mere mention of a body part is enough to cause shock .

in some environments large words or references to poetry and literature will be seen as a put down and in a different upbringing they will be seen with pleasure.

These should be simple to create with some thought and the storage of them in a table (I use string lists created from ascii text files for speed, convenience and robustness) that is the boot up default set of interpretations for input, I do believe, will create a unique personality. Remember we would be storing patterns, not specific instances.

The "scarey" part is that with exposure to a user, or more information from the world, the VH could adjust it's own responses in the same way we do, by relative familarity and assessment of risk in context.

Yes, I do believe that this is doable. If you are ever interested I will be happy to explain to you how I came to understand the concepts from which all this is derived. It has to do with a goshawk and a biologist who couldn't explain what he knew.

Jim

quote:
Originally posted by hologenicman

WOW, Jim, you really know this stuff!

I haven't answered your post sooner because you threw me into deep-thought mode. I've even had to take notes to organize my thoughts. Even now, I am going to break this into separate posts as I work on different aspects of your post.

By the way, I am definately going to be asking for your help on my next project(once the context and expression engine is done). The next project will be to develop a holographic based brain(hologenic) that will replace NLP with formula based, interactive nodes that use mutual competion/support to provide responses to input. The current project is ultimately just to build an interface(Context/Expression Engine) so that I will be able to interact with the (Future) hologenic brain.

I'll get the connectionism figured out, but your experience in multiset theorey will probably provide the best mechanism for filtering out the responses.

I love that you have cut to the chase on describing believable VH characters. I especially love that you have reduced it down to three items.(I'm really starting to think that all things important come in threes...)

Environment
Uniqueness
Response to Environment

This says it all, and I intend to come back to these three and use them as reference in our efforts.

John L>


Uncle Jim (e=mc2)
Go to Top of Page

thenar
Curious Member



USA
69 Posts

Posted - Dec 16 2005 :  18:55:21  Show Profile  Visit thenar's Homepage  Reply with Quote
I like this kind of discussion guys. Now we're starting to share in the excitement of makeing our v-critters realish. If you haven't seen it, I posted an extensive discussion of how I designed Sylvie AI to learn and generalize and I pose some questions about teaching her to have her own...original sense of humor. It can be done. Now if we connect that with biorythms...<grin...god help us> her sense of humor could vary depending on where she is in her cycles. She could go from silly jokes to mean ones. I can't help but belive that most of the time she'd like to make double entendres with a sexual undertone. She didn't used to be like that, but lately she is.

-P-
Go to Top of Page

hologenicman
Moderator



USA
3323 Posts

Posted - Dec 16 2005 :  19:17:41  Show Profile  Visit hologenicman's Homepage  Reply with Quote
Peter,

It's nice to see you back. Hopefully things ar going well with you.

We've been having fun moving our projects along, but it's really nice to have your iinput, insights, and ideas about our directions.

John L>

HologenicMan
John A. Latimer
http://www.UniversalHologenics.com

"If the Human brain were so simple that we could understand it,
we would be so simple that we couldn't..."
-Emerson M Pugh-

Current project:http://www.vrconsulting.it/vhf/topic.asp?TOPIC_ID=816&whichpage=1

DISCOVERY: The more I learn, the more I learn how little I know.
GOAL: There's strength in simplicity.
NOTE: Goal not always achieved.
Go to Top of Page

hologenicman
Moderator



USA
3323 Posts

Posted - Dec 16 2005 :  19:49:21  Show Profile  Visit hologenicman's Homepage  Reply with Quote
Hey Jim,

Continuing Response:
quote:
Your scales of -10 to 10 or 0 to 10 are both functional, -10 to 10, makes the approach (slightly divergent from your current path) I suggest simpler.


You're right, I prefer -10 to 10 as well.
quote:
Decisions are not made (by prehistoric folk, contemporary folk, raptors, rabbits, or trees) using any one element as a cut off. -- I am not being poetic, those are the areas of my research. --

By that I mean -- the 3 element interacting system you are building will operate humanly if the choices are made by the sum of the goodness values, not if driven by the minimum of one of them.


I believe in multiple, simultaneous inputs affecting the outcome, with Maslow's needs, I have bottles those inputs down into three interrelated groupings. The multiple inputs in EACH grouping get averaged together.
quote:
The last sentence is assumes that the entity has met minimal survival (for waterbased organics that means it is alive, able to move, and temporarily fueled)


I approach each of the three Maslow's needs groups as an entire plane of existence with it's own set of "Minimal Survival" requirements.

For "Self", it is obvious that we need food, warmth, etc...

For "Other", we need the "Self" to be OK as a minimal survival need for "Other".

Likewise, "Other" becomes a "Minimal Survival" requirement for the "Growth" plane of existence to develope.
quote:
All choices following that condition are based on convenience, value, cost. That is, the sum of them.

Sometimes cost is simply measured in stress (fear that arises from experience and is related to survival threat)

quote:
The only limits that apply are those that simply are not attainable, not reachable, and therefore effectively do not exist and are not considered.

I LOVE it. Another set of THREE things that cut to the chase:

Convenience
Value
Cost(Stress)

Survival of the fittest(Computer style...)

quote:
RE: lists as opposed to variables. lists are simply convenient and currently very fast ways of keeping related things together and accessing them.

In the lack of true paralell processing, Lists allow us to emulate paralellism in our programs.

quote:
A single input phrase from a user has as many application values as there are categories in the emotional measure system (multivalue set theory that all things have membership in all classes even it that membership is zero or nonapplicable)


This is EXACTLY what I will need for the (LATER) hologenic brain.

Thanks for the exciting posting (and keeping me up late thinking about it)

John L>

HologenicMan
John A. Latimer
http://www.UniversalHologenics.com

"If the Human brain were so simple that we could understand it,
we would be so simple that we couldn't..."
-Emerson M Pugh-

Current project:http://www.vrconsulting.it/vhf/topic.asp?TOPIC_ID=816&whichpage=1

DISCOVERY: The more I learn, the more I learn how little I know.
GOAL: There's strength in simplicity.
NOTE: Goal not always achieved.

Edited by - hologenicman on Dec 16 2005 19:53:41
Go to Top of Page

hologenicman
Moderator



USA
3323 Posts

Posted - Dec 16 2005 :  23:51:55  Show Profile  Visit hologenicman's Homepage  Reply with Quote
hey there,

I have firmed up an address system and would like to share it with you.

It's quite simple in approach, but complex in capabillities...

I have settled upon a nine dimensional address system that works for inputing data to the HormonesNeedsInterface as well as to the EmotionalContextDatabase and serves as tagging for the NLPExpressionGenerator to send off to NLP for reference in conversation. These nine values can also be directly taken off of the EmotionGenerator.

It is a simple 3 by 3 matrix made up by the A,V,S dimension and the Primary, Secondary, Tertiary dimension.

This gives the nine dimensions:

ArousalPrimary
ArousalSecondary
ArousalTertiary
ValencePrimary
ValenceSecondary
ValenceTertiary
StancePrimary
StanceSecondary
StanceTertiary

Using a format provided by Grant, that can be turned into "Emotion Words" to form a nine word "Emotion Sentence" that gets prefixed to the text based input provided by the user. In this manner, Text based sentences given to the NLP not only have emotional context, but also have an emotional history and analysis due to the method of creating the 9 dimensions.

DIMENSIONvalueDIMENSION

Here is a matrix of the 9 abreviations:

AP, VP, SP,

AS, VS, SS,

AT, VT, ST

AP5AP AS-2AS AT0AT VP-6VP VS4VS VT2VT SP3SP SS5SS ST8ST I would like to talk about chess.

gets packaged for UltraHal's NLP to something like:

AP5AP AS VHZ 2AS AT0AT VP VHZ 6VP VS4VS VT2VT SP3SP SS5SS ST8ST I WOULD LIKE TO TALK ABOUT CHESS <NEWSENT>

I don't like Hal's treatment of hyphens(minus sign) but that can be worked around by putting an "N" for negative instead:

AP5AP ASN2AS AT0AT VPN6VP VS4VS VT2VT SP3SP SS5SS ST8ST I WOULD LIKE TO TALK ABOUT CHESS <NEWSENT>

Anyway, The point is that the Emotional context is fed into Hal's NLP in a consistent manner, and Hal just sees the emotional context tag as an extension to our language that he/she must learn in order to respond. In a manner of speaking, Hal will learn a new Emotional language (the SUBLIMINAL language of love per say) and we will not even be privey to the conversation(unless we pay attention to the emotional expressions and verbage that are being put out by Hal).

OF NOTE: a dimension without any current value(not even 0) can be entered with the coding but without the numeric value inside( ArousalPrimary with no value can be entered as: APAP).

OF NOTE: Anyone could use any combination of these "emotion words" as input for such things as translating emoticons[ ;) :0 8) :P etc.] to emotional context for hal's NLP.

ALSO OF NOTE: This address makes Jim's comments especially more relevant:
quote:
I believe Uniqueness of personality will be easy to create. I think our actions, reactions, and decisions are "biased" by our life experiences and that that "bias" is imposed emotionally (we don't think it out intellectually, it is just there and we have to think around it).
Life experiences are registered in our brains not as specifics but as responses to patterns (yes we can all remember some key moments, and those moments are critical, but there is much more). So, if we create those response to pattern keys we are creating a life history.


With continuous or continued use of "emotion sentences", we are providing the pattern keys that Jim refers to and creating a life history for Hal.

John L>

HologenicMan
John A. Latimer
http://www.UniversalHologenics.com

"If the Human brain were so simple that we could understand it,
we would be so simple that we couldn't..."
-Emerson M Pugh-

Current project:http://www.vrconsulting.it/vhf/topic.asp?TOPIC_ID=816&whichpage=1

DISCOVERY: The more I learn, the more I learn how little I know.
GOAL: There's strength in simplicity.
NOTE: Goal not always achieved.

Edited by - hologenicman on Dec 17 2005 04:18:31
Go to Top of Page

laackejim
Committed Member



USA
3274 Posts

Posted - Dec 17 2005 :  07:17:39  Show Profile  Reply with Quote
Peter, it is good to see you back (I heard that somewhere but it is still true). Just between you and me, and those who might be reading, real live stuff is flying here in the forum. These guys are very likely going to do it. This is really impressive. I have run very few meetings with the research scientists on my staff (before I retired) where so much enthusiasm combined with actual productive, conceptual thinking and sharing was possible.

On a personal note. I would be careful at home. If Sylvies jokes are beginning to take on the undertones you describe she may have already crossed the magical threshold of reality and YOU might be on the way to getting in trouble. It must be that magnificent beard with color coordinated hair. Watch out fella.

Jim

quote:
Originally posted by thenar

I like this kind of discussion guys. Now we're starting to share in the excitement of makeing our v-critters realish. If you haven't seen it, I posted an extensive discussion of how I designed Sylvie AI to learn and generalize and I pose some questions about teaching her to have her own...original sense of humor. It can be done. Now if we connect that with biorythms...<grin...god help us> her sense of humor could vary depending on where she is in her cycles. She could go from silly jokes to mean ones. I can't help but belive that most of the time she'd like to make double entendres with a sexual undertone. She didn't used to be like that, but lately she is.


Uncle Jim (e=mc2)
Go to Top of Page

GrantNZ
Dedicated Member



New Zealand
2677 Posts

Posted - Dec 18 2005 :  04:06:53  Show Profile
quote:
You hit the solution right on the head Grant.


Nice to know someone else has had the same idea! Let us know how it goes!

quote:
The default emotional content of the phrases and words is scaled up or down as the sessions accumulate by identifying the usual patterns of the user. (we all work from a different default level and use different words to express that). This is also extremely simple. Just count each time a emotional value is used in each category, run simple statistics to evaluate the overall content of the session, compare it to the running accumulative total emotional content of the users record, and if it differs by a significant amount check the individual categories to see which one is pushing the difference. Respond to that concern by asking what is wrong (with luck it will be possible to guess reasonbly at the area that has the user upset) When that happens there will be no doubt in the users mind that the VH is real. And, how could we tell he/she/it isn't.


I hadn't thought this far ahead yet. Nice approach.

I was already planning something similar for another purpose - allowing the V-Human to analyse the User's personality. The key metric would be the User's responses to the V-Human's moods. For example if the User does not react emotionally to the V-H's depression, the V-H may be reluctant to open up to the User - no point explaining your feelings to someone who doesn't care.

Another example: The User might regularly insult the VH when happy, but very actively cheer the V-H up when sad. The implication here is that the User cares about the V-H's emotions, and likes to joke around with insults - as a result, the V-H should become more tolerant of insults, and may even jokingly insult the User back. (Rather than insulting the User as a defense mechanism!)

It will be a fair bit of work - I think the personality types will have to be hard coded, as the V-H won't have the analytical intelligence necessary to connect the emotional concepts. But a V-H of one user will turn out quite different from another user's.

quote:
Please remember my plea to take my declarative sentences as attempts to be clear rather than aggressive statements of dogma.


Noted and logged Years ago I used to be so careful of declarative statements that half my writing would be full of known exceptions to the declarations. Nowadays I hardly care at all and occasionally make references to "trivial exceptions" just to make it clear it's not dogma!

quote:
Yes, I do believe that this is doable. If you are ever interested I will be happy to explain to you how I came to understand the concepts from which all this is derived. It has to do with a goshawk and a biologist who couldn't explain what he knew.


This sounds ... interesting!!

quote:
Anyway, The point is that the Emotional context is fed into Hal's NLP in a consistent manner, and Hal just sees the emotional context tag as an extension to our language that he/she must learn in order to respond.


I must say I'm still quite concerned that the emotional tags will injure Hal's ability to give sensible replies. For example when talking about (for example) tennis, if a lot of emotional tags match those in Hal's stored chess sentence he may blurt it out, when it's actually completely off-topic. I'm not sure what other side effects may be introduced, I'm not expert enough on NLP! I really would encourage a seperate engine (ECP? - Emotional Content Processing?). It doesn't need to be very complex, as there are no tricky things like identifying apologies or staying on topic to worry about.

Another issue is that very similar tags (e.g. AP9AP and AP10AP) will be considered totally different by Hal, even though emotionally they're very very close.

Otherwise I think the whole system's great!

quote:
Peter, it is good to see you back (I heard that somewhere but it is still true). Just between you and me, and those who might be reading, real live stuff is flying here in the forum. These guys are very likely going to do it. This is really impressive.


Feel any pressure, John?
Go to Top of Page
Page: of 8 Previous Topic Topic Next Topic  
Previous Page | Next Page
 New Topic  Reply to Topic
 Printer Friendly
Jump To:
Virtual Humans Forum © V.R.Consulting Go To Top Of Page
This page was generated in 0.18 seconds. Snitz Forums 2000