hadal 0.0.1

Tool for mining/alignment parallel texts

hadal

hadal /ˈheɪdəl/ is a tool for parallel sentence mining with pretrained models.

🚧🚧🚧The project is under active development. The changes in the API can be breaking. 🚧🚧🚧

Installation

See INSTALL.md for detailed installation instructions.

Implemented methods

Method	Alignment type
margin-based	one-to-one
soon...

Quickstart

See demo file demo.py for more details.

$ python demo.py

# score, source_sentence, target_sentence
[
    (1.5549, "I think I like wine now.", "Je pense que j'aime le vin maintenant."),
    (1.5079, "She eats one apple every day.", "Elle mange une pomme chaque jour."),
    (1.4353, "They serve pizza dogs in the cafeteria.", "Ils vendent des hot-dogs à la cafétéria."),
    (0.4112, "Empty sentence.", "Ce jeu se joue sur le vaisseau spatial."),
]