12 KiB
D&D Name and Language Generator
This package is a fantasy language generator. By defining a number of characteristics about your imagined language -- the graphemes, their relative frequency distributions, the construction of syllables, and so on -- you can generate random but internally consistent gibberish that feels distinct, evocative, and appropriate to your setting.
Usage
The fanlang command-line utility supports three commands:
- names: generate names
- text: generate a paragraph of text
- list: list the supported language in the current language pack
Examples:
% fanlang --language=dwarvish names --count 5
Tiny Châ Pothesadottyr
Khâkhu Zhûdothir
Quiet Ke Vêdothir
Cû Tozhon
Big Pâ Thadottyr
% fanlang --language=dwarvish text
Cû ne do tho khâ tasha, vê wûva lû, ku phu thâ thê, tûko kê, pevo kâ têtetv zha
pataso keks khate? Fâ zhû shû yf pho pa me. Dupha dê thê khâ! Shikm tu! Cê
sâdêto. Dê yo nâ topho, my sû pida phe, vi phûtw châcho, po sotê?
% fanlang list
Abyssal
Celestial
Common
Draconic
Dwarvish
Elvish
Gnomish
Halfling
Infernal
Lizardfolk
Orcish
Undercommon
Language Packs
A Language Pack is a python package that defines one or more language modules. The default language pack includes a number of D&D languages with rules built according to the conventions established by my D&D group over several years of play in our homebrew setting.
The default language pack is language.languages; each submodule contains a README that describes the basic characteristics of the language, along with examples.
Using Your Own Language Pack
You can override fanlang's default language pack by specifying the FANLANG_LANGUAGE_PACK environment variable:
# Create your ancient_elvish module in campaign/language_pack/ancient_elvish
% FANLANG_LANGUAGE_PACK=campaign.language_pack fanlang list
Ancient Elvish
Setting the Default Language
'common' is the default language module. You can override this by setting the FANLANG_DEFAULT_LANGUAGE environment variable:
% FANLANG_DEFAULT_LANGUAGE=gnomish fanlang names --count=1
Jey Lea
You can read about creating custom language packs below.
Library Quick Start
You can load all supported languages in a language pack using language.load_langauge_pack():
>>> import language
>>> language_pack, supported_languages = language.load_language_pack()
>>> common = supported_languages['common']
>>> common.word(2)
['apsoo', 'nirtoet']
>>> common.text()
Proitsiiiy be itkif eesof detytaen. Ojaot tyskuaz apsoo nirtoet prenao.
>>> commoner = f"{common.Name}"
"Quiet" Gushi Murk Lirpusome
>>> common.Name.name()
{
name: ['Gushi', 'Murk'],
surname: ['Lirpusome'],
adjective: ["Quiet"],
}
You can also load individual languages directly:
>>> from language.languages import common
>>> common.Language.word(2)
['apsoo', 'nirtoet']
>>> str(common.Name)
"Quiet" Gushi Murk Lirpusome
Defining a New Language Pack
Language packs are python packages with the following structure:
language_pack:
__init__.py
language_name:
__init__.py
README.md
base.py
names.py
rules.py
...
Language Modules
A language consists of several submodules:
base.py, which contains grapheme definitions and aLanguagesubclass;names.py, which defines theNameGeneratorsubclasses; andrules.py, which is optional, and defines the rules all words in the language must follow.
Read on for a discussion of each of these components.
Language Generators
Let's look at a simple example, the Gnomish language. Here's the Language
subclass defined in base.py:
from language import defaults, types
from .rules import rules
consonants = types.WeightedSet(
("b", 0.5), ("d", 0.5), ("f", 0.3), ("g", 0.3),
("h", 0.5), ("j", 0.2), ("l", 1.0), ("m", 0.5),
("n", 1.0), ("p", 0.5), ("r", 1.0), ("s", 1.0),
("t", 1.0), ("v", 0.3), ("w", 0.2), ("z", 0.1),
)
suffixes = types.equal_weights(["a", "e", "i", "o", "y"], 1.0, blank=False)
Language = types.Language(
name="gnomish",
vowels=defaults.vowels,
consonants=consonants,
prefixes=None,
suffixes=suffixes,
syllables=types.SyllableSet(
(types.Syllable(template="consonant,vowel,vowel"), 1.0),
(types.Syllable(template="consonant,vowel"), 0.33),
),
rules=rules,
minimum_grapheme_count=2,
)
Defining Graphemes
A Language definition includes graphemes, the basic building blocks of any
language. We start with vowels, consonants, which are required in every
language; Gnomish also includes suffixes, but no prefixes. Each
grapheme is a WeightedSet, which is like a regular set except its members
consist of a tuple of a string and a relative weight from 0.0 to 1.0. These
weights will be used when selecting a random grapheme.
Gnomish uses the default vowels, defined in the language.defaults submodule, but define our consonants with a subset of English consonants. By experimenting with different sets and different weights, you can generate radically different feeling text output!
Gnomish also uses suffixes, which are graphemes added to the ends of words.
Here we use the helper function types.equal_weights(), which returns
a WeightedSet where each member is given the same weight. Normally this
function also inserts the grapheme ("", 1.0) into the set, but we disable
this behaviour by specifying blank=False.
Defining Syllables
A syllable is a collection of graphemes, including at least one vowel. When we
create words, we select a random syllable template from a SyllableSet, which
is just a WeightedSet whose members are Syllable instances. Each Syllable
declares a template, and like graphemes, has a weight associated with it that
will make it more or less likely to be chosen for a word.
A syllable's template consists of a comma-separated string of grapheme names.
In Gnomish, we have two possible syllable templates, consonant,vowel,vowel
and the shorter consonant,vowel, which will be selected one third as often.
Templates also support randomly-selected graphemes by joining two or more
grapheme types with a vertical bar, for example vowel|consonant would choose
one or the other; vowel|consonant,vowel would result in a vowel or
a consonant followed by a vowel.
How Words Are Constructed:
The main interface for callers is word(), which returns a randomly-generated
word in the language according to the following algorithm:
- Choose a random syllable from the syllable set
- For each grapheme in the syllable 3. Choose a random grapheme template 4. Choose a random sequence from the language for that grapheme 5. Validate the word against the language rules
- Repeat 1-5 until a valid word is generated
- Add a prefix and suffix, if they are defined
When graphemes are chosen, the following rules are applied:
- Every syllable must have at least one vowel; and
- A syllable may never have three consecutive consonants.
A More Complex Example
The Common language is a more complex definition, with language-specific prefixes, suffixes, vowels, and consonants, and many possible syllables. This results in highly-varied text.
Rules
Rules are a set of callables that accept a language instance and a word. The callable returns True if the word passes some test, and False otherwise. Every randomly-generated word must pass all defined rules for the language, or it is rejected.
[The language.rules module(language/rules.py) contains a number of useful rules that are applied by default to most languages, mostly used to aid readability and generate words that are pronouncable. Here's a simple example:
def too_many_consonants(language: Language, word: str) -> bool:
found = re.compile(r"[bcdfghjklmnpqrstvwxz]{3}").findall(word)
if found == []:
return True
logger.debug(f"{word} has too many contiguous consonants: {found}")
return False
This rule ensures that a word does not contain more than 3 english consonants in a row.
Defining Language-Specific Rules
Rules are passed as a set of callables to the Language constructor, so they can be anything you want, defined anywhere you want. By convention, language packs use a separate rules module when building custom rule sets.
Name Generators
Name generators are similar to Language generators, but with a few key differences. Here is a simple example, also from the Gnomish language:
from language import types
from language.languages.gnomish import Language
Name = types.NameGenerator(
language=Language,
templates=types.NameSet(
(types.NameTemplate("name,surname"), 1.0),
),
)
NobleName = Name
In Gnomish, names are straightforward, consisting of a name and a surname, and there is no distinction between regular names and the names of the nobility. By contrast, Elvish names are complex, consisting of multiple distinct parts, including place names, affixes, and separate rules for common and noble name construction.
Defining Names
Name Generators are defined with a Language, and one or more NameSet templates. NameSets are equivalent to SyllableSets, but instead of creating templates for the construction of syllables from sequences of graphemes, they define sequences of parts of names -- names, surnames, titles, nicknames, and so on. They follow the same semantics as Syllables, allowing for a large variety of potential names.
By default, both names and surnames are generated automatically by calling NameGenerator.language.word(). Thus, the simplest name generator will simply follow all the rules of the language itself and generate one or more random words. You can override multiple aspects of a language's rules for word generation by passing NameGenerator additional arguments. For example, here is a generator for the names of locations in the Elvish language:
from language import types, defaults
from language.languages.elvish import Language
PlaceName = types.NameGenerator(
language=Language,
syllables=types.SyllableSet(
(types.Syllable(template="vowel,vowel|consonant,vowel|consonant"), 1.0),
(types.Syllable(template="consonant,vowel|consonant,vowel|consonant"), 0.3),
),
templates=types.NameSet(
(types.NameTemplate("affix,name"), 1.0),
),
affixes=types.WeightedSet(("el", 1.0)),
adjectives=defaults.adjectives,
)
Note how we declare new syllables, affixes, and adjectives, which will replace the Language's default behaviours.
Subclassing
Sometimes we need even more control than providing new syllable sets and weighted sets for graphemes. Subclassing NameGenerator gives you significant control over how names are constructed. For example, Elvish surnames are based on Elvish place names; to accomplish this, we subclass NameGenerator and override the method used to generate surnames:
class ElvishNameGenerator(types.NameGenerator):
def __init__(self):
super().__init__(
language=Language,
syllables=Language.syllables,
templates=types.NameSet(
(types.NameTemplate("name,affix,surname"), 1.0),
),
affixes=types.equal_weights(["am", "an", "al", "um"], weight=1.0, blank=False),
adjectives=defaults.adjectives,
titles=defaults.titles,
suffixes=suffixes,
)
self.place_generator = PlaceName
def get_surname(self) -> str:
return self.place_generator.name()[0]["name"][0]