Rule Specification

Selectors

Vowel=aeiou
In another rule, Vowel will be replaced by one of a, e, i, o, or u. The names can be just about anything, except for numbers (which have a special meaning, see below). Be careful with the names, though, as anything matching will be assumed to be a reference. For example, if you make a rule a=q, any time you put a in another rule, it will actually output q.
ProbI=a25b30c45
Example of probabilistic selectors. Default base probability is 100, so in this case a will be chosen 25% of the time, b will be chosen 30% of the time, and c 45% of the time.
ProbII=c50
This is incompletely specified, so ProbII will be c half the time, and half the time it will be blank.
ProbIII=a25bc
If there are choices without probabilities, they split the remainder of the probability. In this case, a has a 25% chance, while b and c both have 37.5% (75%/2).
ProbIV=12a2b3
Finally, you can change the default base probability by putting a number right after the =. a will be chosen 2/12 times, b 3/12 times, and nothing the rest of the time.

Mergers

Syllable:ConsonantVowel
Syllable will be replaced by Consonant AND Vowel (in that order). For example, if Consonant=tpk and Vowel=aeiou, some possible Syllables would be ta, po, ki.
CVmaybeC:ConsonantVowelConsonant50
Probabilities in mergers work a bit differently. If there is a number, it represents the preceding element's chance of inclusion. In this case, the second Consonant has a 50% chance of appearing; example outputs from CVmaybeC are tat, po, kip, kot, and pi.
Repvowel:ConsonantVowel{100,50,...}
You can also specify probability lists with mergers. This allows repetition of the preceding element. The Vowel here will definitely show up once (100%), then has a continuous 50% chance of repeating (due to the ...). The continous chance is optional; if this were just {100,50} then the Vowel would show up once for sure, and twice with 50% likelihood.

Mimics

Eng<[this is a list of words]

In another rule, Eng will be replaced with something a bit complicated. In short, it attempts to replicate the style of the words in the list. Simplest way to use it would be to have a list of words (as above), and then output the rule right after (e.g. >Eng).

In long, it builds weighted trees for each letter, so if you have a list like [that than thanks], if an a is chosen at some point, then 1/3 times it will be followed by n, and 2/3 times it will be followed by n. Longer word lists will create more complex words, though too many might not produce anything useful.

The initial and terminal letters are somewhat special; if all words begin with t in your list, then all words in the output will as well. Similarly, if all list words end with an e, then all output words will, too.

Other

si~shi
Substitution. If "si" shows up in any output, replace it with "shi". Be careful with order, as they will be applied sequentially. If you have ty~ch and then hu~fu, tyu will become cfu.
>Syllable{100,95,90,70,50,...}
Output. Really a special case merger; see the Merger section for details on syntax.

Changes