Kinds (common nouns)

Kinds of entities are expressed as common¹ nouns like cat, book, chair, or candy factory.  Notice from the latter that Imaginarium is okay with a “noun” consisting of a series of words.  That lets you say things like candy factory without having to teach it separately about what candy means, irrespective of factories, and factory, irrespective of candy.

For the most part, you can teach the system a new kind (a new noun) by just using it in a statement someplace.  For example, if you say:

Chairs are wooden, steel, or plastic.

Then it will figure out that chair is a noun, even if you haven’t previously used the word chair.²  Kinds are important because the other sorts of words tend to get organized around them.  For example, you don’t tell it fluffy is an adjective, you tell it:

Cats can be fluffy.

Which tells it that it needs to worry about entities being fluffy when those entities are cats.  But if you told it about some other kind of entity, like books, it doesn’t need to worry about whether they’re fluffy unless for some reason you also specifically told it: books can be fluffy.

Nouns can have multiple words

Imaginarium allows nouns to have more than one word, such as in the example candy factory, above.³  This is really useful, but it’s also a little complicated.  How does the system know that candy factory is a two-word noun, but black cat is a one-word noun with an adjective attached?  The answer is that it always assumes something is a multiword noun, unless you’ve already taught it that the words inside the noun have other meanings.  So if you say:

A cat can be black.
Black cats are great.

Then the first line teaches it that cat and black are separate concepts, and the second that when you have a cat who is specifically black then that cat will always be great.  But if you just say the second sentence, without ever having used the words black or cat, then it will assume that black cat is a single, indivisible noun and subsequent uses of cat, without black, will confuse it.

Subkinds and superkinds

You can teach the system that one kind is a subkind of another kind by saying:

noun is a kind of noun.

For example, chair is a kind of furniture or sword is a kind of weapon.  This tells it that all swords are also weapons.  This can get tedious when you have to teach it about a lot of kinds, so you can just say:

Sword, gun, and club are kinds of weapon.

And it will treat it as equivalent to typing three separate is a kind of statements.  Note that you have to use the commas here so that it doesn’t think there’s a kind of weapon called a sword gun, although if you want that, take out the comma.

Note that when you tell the system to generate an entity of some kind that has subkinds, it will always choose a specific subkind.  That is, if you say “imagine a weapon”, it will always generate a sword, gun, or club, not some generic thing described only as a weapon.  It will generate the different subkinds with equal probability.  But if you want one to be more common than the others, you can put how much more common in parentheses:

Sword (2), gun, and club are kinds of weapon.

This says that swords are twice as common as guns and clubs in our world.  If you want guns to be rare, you can put in a fractional number:

Sword, gun (0.1), and club are kinds of weapon.

Endnotes