kotki 0.4.5

Translate text locally on your machine.

kotki

This package provides Python bindings for kotki, the Bergamot-based translation engine. Fast & easy language translations inside an easy to install Python module.

import kotki

kotki.loadRegistry("/home/user/example/registry.json")

kotki.translate("Whenever I am at the office, I like to drink coffee", "ende")
'Wann immer ich im Büro bin, trinke ich gerne Kaffee'

kotki.translate("Румънците получиха дълго чакани новини: пенсиите и минималната заплата ще бъдат увеличени от 2023 г.", "bgen")
'Romanians have received long-awaited news: pensions and minimum wages will be increased from 2023'

kotki.translate("jij bent geboren in de stad Den Haag.", "nlen")
'You were born in The Hague.'

Requirements

Lets grab some requirements before we do pip install.

apt install -y ccache rapidjson-dev cmake libpcre++-dev libpcre2-dev python3-dev pybind11-dev

Get MKL installed. For example, this does the installation on Ubuntu 21.

wget -qO- 'https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS-2019.PUB' | sudo apt-key add -
sudo sh -c 'echo deb https://apt.repos.intel.com/mkl all main > /etc/apt/sources.list.d/intel-mkl.list'
sudo apt update
sudo apt install -y intel-mkl-64bit-2020.4-912

In case you cannot find the correct version to install, you may use the package manager to search for the correct package name: apt search intel-mkl-64bit-2020

pip install kotki

at which point you can do import kotki inside your Python application.

API

The API is straight-forward and contains only 3 functions:

• loadRegistry(path), read available translation models
• translate(text, language), translate some text
• listModels(), list available translation models

Models

The translation models that kotki uses are 'borrowed' from the Mozilla Firefox Translations extension. You need to manually download these models. They are conveniently packaged into a single archive that can be downloaded at github.com/kroketio/kotki/releases.

registry.json is included in this archive - which is needed for the loadRegistry() call.

Performance / footprint

Translations are fast - (probably) faster than other Python packages that do language translation. Translating a simple sentence is usually under 10ms (except the first time, due to model loading). Loading a single translation model seems to take up around 40MB in RAM.

Translation models are loaded on-demand. This means that model loading does not happen during loadRegistry() but during the first use of translate() - which typically takes (only) 100ms (per model). So if you have a project that uses both translate('foo', 'enfr') and translate('foo', 'fren') - you'll be using 2 models (and consequently 80MB worth of RAM during the duration of your program).

Note that translations are done synchronously (and thus are 'blocking'). If you need an async/callback style approach, look at the Bergamot-Translator.

Acknowledgements

This project was made possible through the combined effort of all researchers and partners in the Bergamot project (Jerin Philip, et al). The translation models are prepared as part of the Mozilla project. The translation engine used is bergamot-translator which is based on marian.

License

MPL 2.0