The Middle of the Story

It is interesting. You the readers are coming in on the middle of the story. I have been working on this stuff for about 20 years, and so any time I have new developments to report, they exist in my mind in a context of much work having been done before. However you are coming in in the middle of the story, so it is catch as catch can whether you will understand what I am saying at one time or another.

Understand what you can. Don't worry about what you can't. It's probably because I have not provided enough context. I provide as much context as I can bear for every article, and try to put a little more in later as I read the articles. If you have questions or do not understand something, I would very much appreciate your comments or questions. Thank you.

[0] comments (389 views) |  [0] Trackbacks   [0] Pingbacks

A Model of English that a Computer Can Understand - Part II

When making a Model so that a computer can understand English, one thing you must keep in mind as you work, is to come up with parameters that a computer can understand. The approach I take is categorization with the goal of creating a Rule of Twenty Profile. You can categorize words to your hearts content, but you have to remember to make the categories things that a computer can process. So as I come up with charateristics for the profile, I need to make sure a computer can use them.

For example, I am presently building a 'plural' category/characteristic.:

1. I started by ignoring plurals in my part of speech model.
2. Then I tried to add a 'quantity' magnitude characteristic to the rule of twenty profile with 20 different possible, but this did not work because it was too many possible values,
3. then I tried to add a 'plural' characteristic with only two possible values
4. now I have expanded it to five possible values: plural(4), almost plural(3), pair(2), single(1), zero(0). This looks the best of all of my attempts so far.

I am using the 'almost plural' value to contain words that are part of collections of words for example the months of the year: January February march April etc. And also for singular measures for example 'meter' whcih gets a 3 as opposed to 'meters' which gets a 4. I hope this is something a computer can use.

OK, now I'm thinking of changing substances from plural (4) to almost plural (3).

[0] comments (413 views) |  [0] Trackbacks   [0] Pingbacks

A Rule of Twenty for the English Language

I have this crazy idea that I can apply the Rule of Twenty Profile System to create a profile for every English word, and therefore have a computer generated character be able to both understand and compose dialog. Actually, I used the Rule of Twenty with D&D spells once so I know that a fairly high level of conceptual complexity can be managed (then the system collpased into six variables showing that there wasn't much to the spell system). I don't work with spells any more since they are evil, but I think that this effort with English has a tiny slight chance of working.

If the chance of succeeding is for example 1% then it's worth a shot. Even if it doesn't work I'll be able to demonstrate how I put together a rule of twenty system.

I have been beating up on a group of about 5000 common words for four months now. The first thing I did was set up a four bit binary system to identify part of speech. Here are the results of my first two phases:

[0] comments (437 views) |  [0] Trackbacks   [0] Pingbacks

The Field of Artificial Creativity

About 20 years ago when I first started thinking about Artificial Creativity as a field, I produced the following diagram to describe it:

[0] comments (420 views) |  [0] Trackbacks   [0] Pingbacks

Creativity at Work

One approach to Artificial Creativity is to have the users of software do the creating. For reason which I have touched on and hope to focus an article on some time, I think it's important to have as many people as possible involved in making creative systems. I think the users of computer systems should do it, but to have users do artificial creativity requires two needs:

1. which I touched on in my Algorithmic Sublanguages article is to have programs written so that users can affect the algorithms.
2. The other is to have users interested in creating to begin with.

So this question here is where might users want to create. There are many many places, but here is an opening list. Here is where users might want to be creative:

• Data Processing and Input:
  • creative in organizing data for input
  • creative in organizing data for analysis
  • creative in determining what reports are needed
  • creative in identifying how to change a program to improve its throughput
  • creative in coming up with data products that may serve downstream people


• General Business:
  • creative in writing memos
  • creative in determining which memos to write
  • creative in coming up with new passwords
  • creative in coming up with search terms


• Software Development:
  • creative in writing programs
  • creative in designing programs
  • creative in refactoring programs
  • creative in coming up with test cases for programs
  • creative in naming variables, methods, classes, tables and fields


• Management:
  • creative in organizing people
  • creative in organizing group hierarchies
  • creative in identifying which tasks to give to which employees
  • creative in conceiving tasks that the group needs to do
  • creative in designing workflows


• Network Administration:
  • creative in coming up with things to try to fix a problem
  • creative in designing networks


• Web Page Design:
  • creative in coming up with new ideas for web page layout
  • creative in coming up with new ideas for web site architecture
  • creative in coming up with color schemes
  • creative in coming up with menu labels


• Artificial Creativity:
  • creative in coming up with new models
  • creative in coming up with new things to model
  • creative in coming up with new modeling engines
  • creative in designing model interpreters


• Sales:
  • creative in writing ad copy
  • creative in coming up with sources for new prospective customers

These are all possible areas for exposing the algorithms in your programs to users. What I want is to find where artificial creativity could be used professionally. This is a pretty good set of starting places. In addition, if you are just plain interested in where creativity can be used at work, this is a starting list.

[0] Trackbacks   [0] Pingbacks

The Terrain Generator

Here is my design for a terrain generator. Artificial Worlds need terrain. This terrain generator design is not finished but the part that is finished I can show. The world is divided into a cube of six sides. Each side is covered with regions of terrain. Each region has a center and a size. There are a few large regions like the poles that have a very large size. Most regions are smaller - continents, islands, oceans, forests, mountainous areas etc. The data kept for each region is size, type of terrain, game significance. I will leave off game significance for now. Regions overlap producing complex terrain.

Letter	Terrain	'Opposite'	Color
A	Arctic	Tropical	whiTe
B	Bedrock	Eolian	Neutral gray
C	Cities / Civilized	Hunters	blacK
D	Desert / Dry	Rainy	Orange
E	Eolian (dirt)	Bedrock	Faded brown
F	Forest	Q-swamp	Jaded Jade
G	Green / Fertile	Infertile	Green
H	Hunters	Civilized	Saturated Salmon
I	Infertile	Green	Purple
J	Jungle / Thicket	Open	Quixotic chartreuse
K	K-farms / farmers	Nomads	Murky Mustard
L	Lake / Fresh water	Sea	Achromatic Aqua
M	Mountain	Plain	Red
N	Nomads	K-farms	Cold Cyan
O	Open	Jungle	Dark olive
P	Plain	Mountain	yEllow
Q	Q-swamp	Forest	Infuscated Indigo
R	Rainy / Rainforest	Desert	Light Lavender
S	Sea / Salt water	Lake	Blue
T	Tropical	Arctic	Hot pink

This system is from the Rule of Twenty; The colors definitely are. Because more advanced societies can substantially affect terrain, I need to include the four lifestyle types: farming, nomads, hunting, and cities. Opposites are not exactly oppoiste. For example bringing Lake and Sea together gives you an estuary, and bringing Mountain and Plain together gives you a plateau or a mesa etc. By putting two or more of these together you can create a large number of different types of terrain. This will give a pretty colorful map, but hopefully since I have made many of the larger terrain types have fairly normal map colors, it will look reasonable.

The terrain types will gather together in profiles based on the size of each region and how far the pixel is from the center of each. This using the Rule of Twenty determines the color.

[0] comments (423 views) |  [0] Trackbacks   [0] Pingbacks