Formality Classifier

A multilingual classifier trained to predict the formality of text for the language pairs: EN-RU, EN-KO, EN-VI, and EN-PT. The xlm-roberta-base model was fine-tuned on human-written formal and informal text following the setup from Briakou et al., EMNLP 2021. Please note that the models come with the following licensing.

EN-HI

Updated EN-HI pre-trained model using updated sacremoses tokenizer addressing #3. Model was trained on open datasets using the Sockeye 3 PyTorch NMT toolkit. For further details see Pre-trained models. Please note that the model comes with the following licensing.

EN-IT,RU

Pre-trained baseline models

Baseline models for EN-IT and EN-RU language arcs. Each model is trained on open datasets using the Sockeye 3 PyTorch NMT toolkit. For further details see Pre-trained models. Please note that the models come with the following licensing.

EN-DE,ES,JA,HI

Pre-trained baseline models

Each model is trained on open datasets using the Sockeye 3 PyTorch NMT toolkit. For further details see Pre-trained models. Please note that the models come with the following licensing.