Small Python script to generate simple sentences for a synthetic language with minimal concatenative morphology

generate_sentences_with_morphology.py

import random
import bisect
from collections import Counter


class DiscreteProbabilities:
    def __init__(self, ps):
        keys, values = zip(*sorted(ps.items(), key=lambda kv: kv[1], reverse=True))
        self.values = numpy.cumsum(values)
        self.values /= self.values[-1]
        self.keys = keys

    def generate(self):
        x = numpy.random.random()
        i = bisect.bisect(self.values, x)
        return self.keys[i]


def phonemes():
    """Generate the phoneme inventory of a simlang"""
    return {
        "c": DiscreteProbabilities(
            {
                k: 1 / i
                for i, k in enumerate(
                    [
                        "p",
                        "b",
                        "t",
                        "d",
                        "c",
                        "k",
                        "g",
                        "'",
                        "f",
                        "s",
                        "h",
                        "m",
                        "n",
                        "ŋ",
                        "w",
                        "l",
                        "r",
                        "j",
                    ],
                    1,
                )
            }
        ),
        "v": DiscreteProbabilities({"a": 0.5, "ə": 0.33333333, "i": 0.25, "u": 0.2}),
    }


def morphemes(inventory):
    nouns = DiscreteProbabilities(
        {
            (
                inventory["c"].generate(),
                inventory["v"].generate(),
                inventory["c"].generate(),
                inventory["v"].generate(),
            ): 1.0
            / i
            for i in range(1, 1001)
        }
    )
    noun_affixes = DiscreteProbabilities(
        {
            (inventory["c"].generate(), inventory["v"].generate()): 1.0 / i
            for i in range(1, 11)
        }
    )
    verbs = DiscreteProbabilities(
        {
            (
                inventory["c"].generate(),
                inventory["v"].generate(),
                inventory["c"].generate(),
                inventory["v"].generate(),
            ): 1.0
            / i
            for i in range(1, 1001)
        }
    )
    verb_affixes = DiscreteProbabilities(
        {
            (inventory["c"].generate(), inventory["v"].generate()): 1.0 / i
            for i in range(1, 11)
        }
    )
    other_lexicon = DiscreteProbabilities(
        {
            (inventory["c"].generate(), inventory["v"].generate()): 1.0 / i
            for i in range(1, 101)
        }
    )
    other_grammar = DiscreteProbabilities(
        {
            (inventory["c"].generate(), inventory["v"].generate()): 1.0 / i
            for i in range(1, 101)
        }
    )
    return {
        "n": nouns,
        "ng": noun_affixes,
        "v": verbs,
        "vg": verb_affixes,
        "o": other_lexicon,
        "og": other_grammar,
    }


def sentence(morphemes, g=Counter(), l=Counter()):
    s = ()
    for i in range(numpy.random.randint(3, 10)):
        segment = ["o", "o", "o", "n", "n", "v"][numpy.random.randint(6)]
        form = morphemes[segment].generate()
        if numpy.random.random() < 0.5:
            morph = morphemes[segment + "g"].generate()
        else:
            morph = ()
        s = s + form + morph + (" ",)
        g.update(morph)
        l.update(form)
    return "".join(s) + ".", g, l


inventory = phonemes()
morphemes = morphemes(inventory)
for i in range(130):
    s, g, l = sentence(morphemes)
    print(s)
print(g)
print(l)
print(sum(g.values()) + sum(l.values()))