LANG files are written in a little lightweight markup language for statistical linguistics.
Any basic LANG file can be divided in two parts:
- Groups of phonemes (each phoneme has a weight).
- A grammatical expression.
These files can be interpreted to generate random words.
All phonemes must be grouped into logical units. Each phoneme has its own statistical weight.
For example, here's a group of vowels:
symbols for a vowel:
a: 50
e: 60
i: 40
o: 45
u: 40
Anything between the keyword "for" and the last word is ignored, so these forms are all allowed:
symbols for vowel:
symbols for each vowel:
symbols for any vowel:
You can also use "=" instead ":" when defining a phoneme.
A phoneme's statistical weight gives a priority to the symbol. It must be an integer greater than 0 and less than 100.
Replacements are a way to add exceptions, so they are simple ways to avoid complicated expressions.
All replacements are applied at the level of each word in order.
replacements:
a: aj
ew: ej
The previous replacements would convert
words like vakew to vajkej.
Each LANG file has an expression at the end.
The expression most be at the end, starting in the next line, after:
expression:
An expression is composed of operators that may be nested.
-
Maybe(weight, A): A symbol from the A SymbolSet may appear or not (empty string). The weight states the possibility of A to appear.
-
Maybe(A): Same previous operator, but with default 50 weight.
-
Or(weight, A, B): A symbol from SymbolSet A or B may appear. The weight specifies the possibility of a symbol from A for appearing, and the complement (100 - A) must be the possibility of a symbol from B to appear.
-
Or(A, B): Same previous operator, but with default 50 weight.
-
A + B: The plus sign to append A and B results.