mlotsawa

Description

The purpose of this module is user-friendly translation and transliteration of Classical Tibetan.

Currently, separate classes are provided for translation and transliteration tasks. Both classes will accept either a list of strings, or a single string. The output of each will be of the same type as the input.

You can also run the translator and transliteration functions through a web-based user interface. The web interface will open automatically when you use the code below. You can use this to serve the app to others on your network, as well.

Under the hood, this module uses the T5 transformer architecture, custom fine-tuned on data from Lotsawa House.

The models and datasets used by this project can be found on Hugging Face where you can find more information on training, data collection, and how to use these models and datasets for your own projects.

This module is part of the larger MLotsawa project for machine translation of Literary Tibetan. The code for the entire project, including this module is available on GitHub here.

Installation

pip install --upgrade mlotsawa

Examples

For transliteration:

from mlotsawa.transliterator import Transliterator 

tibetan_text = '<your Tibetan text>'

transliterator = Transliterator()

phonetics = transliterator.transliterate(tibetan_text)

For translation:

from mlotsawa.translator import Translator 

tibetan_text = '<your Tibetan text>' # may be in Tibetan script or phoneticized

translator = Translator()

translation = translator.translate(tibetan_text)

To serve the Web based user-interface:

from mlotsawa.webui import WebUI

webui = WebUI()

webui.run()

The WebUI.run() method has several optional parameters. These parameters and their default values are shown below.

from mlotsawa.webui import WebUI

webui = WebUI()

webui.run(
    logging=False, # keep log of when translations are done through your host
    log_filepath='mlotsawa_logs.txt', # filepath and filename to be used for the log file (i.e. '/my_log_folder/my_logs.txt)
    port=5000, # which port to run the webui from
    autolaunch=True # automatically start the webapp in your default browser
)

License

Shield:

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

Owner

@billingsmoore