Skip to content

Word Level Pre-trained Models

We have released several pre-trained TransQuest models on word-level quality estimation and they are available on Hugging Face Model Hub. We will be keep releasing new models. So please keep in touch.

Available Models

Language Pair NMT/SMT Domain Algorithm Model Name and Link to HUgging Face
English-German NMT IT MicroTransQuest TransQuest/microtransquest-en_de-it-nmt
NMT Wiki MicroTransQuest
SMT IT MicroTransQuest TransQuest/microtransquest-en_de-it-smt
English-Latvian NMT Life Sciences MicroTransQuest TransQuest/microtransquest-en_lv-pharmaceutical-nmt
SMT Life Sciences MicroTransQuest TransQuest/microtransquest-en_lv-pharmaceutical-smt
English-Czech SMT IT MicroTransQuest TransQuest/microtransquest-en_cs-it-smt
German-English SMT Life Sciences MicroTransQuest TransQuest/microtransquest-de_en-pharmaceutical-smt
English-Chinese NMT Wikipedia MicroTransQuest
English-Russian NMT IT MicroTransQuest
*-* Any Any MicroTransQuest

You can load any of the above models with the below code. The full notebook is available here. Let's consider loading microtransquest-en_lv-pharmaceutical-nmt model.

1
2
3
4
5
from transquest.algo.word_level.microtransquest.run_model import MicroTransQuestModel
import torch

model = MicroTransQuestModel("xlmroberta", "TransQuest/microtransquest-en_lv-pharmaceutical-nmt", labels=["OK", "BAD"], use_cuda=torch.cuda.is_available())
source_tags, target_tags = model.predict([["if not , you may not be protected against the diseases . ", "ja tā nav , Jūs varat nepasargāt no slimībām . "]])

Note

* denotes any language. (*-* means any language to any language)