Skip to content

Viterbi-based accelerated tokenizer (Python wrapper)

License

Apache-2.0, MIT licenses found

Licenses found

Apache-2.0
LICENSE-APACHE
MIT
LICENSE-MIT
Notifications You must be signed in to change notification settings

daac-tools/python-vibrato

Repository files navigation

🐍 python-vibrato 🎤

Vibrato is a fast implementation of tokenization (or morphological analysis) based on the Viterbi algorithm. This is a Python wrapper for Vibrato.

PyPI Build Status Documentation Status

Installation

Install pre-built package from PyPI

Run the following command:

$ pip install vibrato

Build from source

You need to install the Rust compiler following the documentation beforehand. vibrato uses pyproject.toml, so you also need to upgrade pip to version 19 or later.

$ pip install --upgrade pip

After setting up the environment, you can install vibrato as follows:

$ pip install git+https://github.com/daac-tools/python-vibrato

Example Usage

python-vibrato does not contain model files. To perform tokenization, follow the document of Vibrato to download distribution models or train your own models beforehand.

Check the version number as shown below to use compatible models:

>>> import vibrato
>>> vibrato.VIBRATO_VERSION
'0.5.1'

Examples:

>>> import vibrato

>>> with open('tests/data/system.dic', 'rb') as fp:
...     tokenizer = vibrato.Vibrato(fp.read())

>>> tokens = tokenizer.tokenize('社長は火星猫だ')

>>> len(tokens)
5

>>> tokens[0]
Token { surface: "社長", feature: "名詞,普通名詞,一般,*" }

>>> tokens[0].surface()
'社長'

>>> tokens[0].feature()
'名詞,普通名詞,一般,*'

>>> tokens[0].start()
0

>>> tokens[0].end()
2

Note for distributed models

The distributed models are compressed in zstd format. If you want to load these compressed models, you must decompress them outside the API.

>>> import vibrato
>>> import zstandard  # zstandard package in PyPI

>>> dctx = zstandard.ZstdDecompressor()
>>> with open('tests/data/system.dic.zst', 'rb') as fp:
...     with dctx.stream_reader(fp) as dict_reader:
...         tokenizer = vibrato.Vibrato(dict_reader.read())

License

Licensed under either of

at your option.