ruccent 1.0.2

Pure Python, no dependencies, simple word accent setter, based on ngram model.

Simple word accent setter.

Pure Python, no dependencies. Based on ngram model.

Trained for the Russian language, on a corpus of ~1000000 word forms, taking into account their frequency in real texts. Accuracy ~97-98%. Speed ~10000 words per second.

from ruccent import Accent

accenter = Accent()
accenter.predict('пробойник')  # -> 4

You can train the model yourself for your language or corpus.

from ruccent import Accent, AccentTrain

# iterable of tuple (accent position, weight/frequency, word)
dataset = [(3, 10, 'begin'), (1, 1, 'happy')]
train = AccentTrain()
train.fit(dataset, ngram=(3,7))
train.save('path/yourself.model')

accenter = Accent('path/yourself.model')
accenter.predict('begin')