RUAccent Python Library

You can install the ruaccent library using standard pip or directly from the GitHub repository.

# Using pip
pip install ruaccent

# Using Git
pip install git+https://github.com/Den4ikAI/ruaccent.git

To use RUAccent, instantiate the RUAccent class, call load() to initialize the models, and then use process_all(text) to apply accents to a string.

from ruaccent import RUAccent

accentizer = RUAccent()
# Initialize with specific settings
accentizer.load(omograph_model_size='turbo3.1', use_dictionary=True, tiny_mode=False)

text = 'на двери висит замок.'
# Process the text
print(accentizer.process_all(text))

The load() method initializes the accentuation engine. You can configure model size, dictionary usage, custom dictionaries, hardware acceleration, and working directories.

Hardware Requirements:

• Minimum 512 MB RAM is required for stable operation (using the tiny omograph model).
• For CUDA support, you must install onnxruntime-gpu and have CUDA configured on your system.

accentizer.load(
    omograph_model_size='turbo2', 
    use_dictionary=True, 
    custom_dict={}, 
    device="CPU", 
    workdir=None
)

The following parameters are available for the load() method:

The process_yo method specifically handles the disambiguation of words involving the letter 'ё' (yo). It uses a specialized model to determine if a word should be pronounced with 'ё' or 'е' based on context.

from ruaccent import RUAccent

accentuator = RUAccent()
accentuator.load()

text = "Все были рады встрече."
# This will resolve whether 'все' is 'всë' or 'все'
result = accentuator.process_yo(text)
print(result)

The process_all method is the primary entry point for accentuating text. It handles sentence splitting, omograph resolution (disambiguating words with multiple pronunciations), and applying stress marks (e.g., +).

Parameters:

• text (str): The input Russian text to process.
• skip_regex (str, optional): A regular expression string. Any text matching this pattern will be skipped and left unaccentuated. This is useful for protecting specific entities or technical terms.

Return Value:

Returns the processed string with stress marks inserted.

from ruaccent import RUAccent

accentuator = RUAccent()
accentuator.load()

text = "Мама мыла раму."
# You can also skip specific patterns
result = accentuator.process_all(text, skip_regex=r'\d+')
print(result)

The ruaccent package provides the RUAccent class, which is the primary interface for accentuating Russian text. You can import it directly from the top-level package.

from ruaccent import RUAccent

The split_sentences function provides a basic sentence splitting mechanism using a regular expression (SENTENCE_SPLITTER). It is less sophisticated than split_by_sentences as it does not account for abbreviations or context-aware joining logic; it simply finds patterns ending in ., ?, !, or ….

Returns:

• List[str]: A list of stripped strings found by the regex pattern.

from ruaccent.text_split import split_sentences

text = "Hello! How are you?"
sentences = split_sentences(text)
# Output: ['Hello!', 'How are you?']

The library uses these predefined sets to handle Russian and English abbreviations during sentence splitting.

JOINING_SHORTENINGS = {
    'mr', 'mrs', 'ms', 'dr', 'vs', 'англ', 'итал', 'греч', 'евр', 'араб', 'яп', 'слав', 'кит',
    'тел', 'св', 'ул', 'устар', 'им', 'г', 'см', 'д', 'стр', 'корп', 'пл', 'пер', 'сокр', 'рис'
}

SHORTENINGS = {
    'co', 'corp', 'inc', 'авт', 'адм', 'барр', 'внутр', 'га', 'дифф', 'дол', 'долл', 'зав', 'зам', 'искл',
    'коп', 'корп', 'куб', 'лат', 'мин', 'о', 'обл', 'обр', 'прим', 'проц', 'р', 'ред', 'руб', 'рус', 'русск',
    'сан', 'сек', 'тыс', 'эт', 'яз', 'гос', 'мн', 'жен', 'муж', 'накл', 'повел', 'букв', 'шутл', 'ед'
}

PAIRED_SHORTENINGS = {('и', 'о'), ('т', 'е'), ('т', 'п'), ('у', 'е'), ('н', 'э')}