Dataset Viewer
Full Screen Viewer
Full Screen
The viewer is disabled because this dataset repo requires arbitrary Python code execution. Please consider
removing the
loading script
and relying on
automated data support
(you can use
convert_to_parquet
from the datasets
library). If this is not possible, please
open a discussion
for direct help.
YAML Metadata
Warning:
empty or missing yaml metadata in repo card
(https://huggingface.co/docs/hub/datasets-cards)
EC40 MNMT Dataset
GitHub: https://github.com/Smu-Tan/ZS-NMT-Variations/tree/main
EC40 is an English-Centric Multilingual Machine Translation Dataset. It has over 60 Million sentences including 40 Languages across 5 Language Families.
Note: The dataset is cleaned and pre-processed using tools like Moses, for more details, please refer to the paper.
Features:
- We carefully balanced the dataset across resources and languages by strictly maintaining each resource group containing 5 language families and each family consists of 8 representative languages.
- EC40 covers a wide spectrum of resource availability, ranging from High(5M) to Medium(1M), Low(100K), and extremely-Low(50K) resources.
- In total, there are 80 English-centric directions for training and 1,640 directions (including all supervised and ZS directions) for evaluation.
- We make use of Ntrex-128 and Flores-200 as our validation and test set.
Languages and Family
Family | Languges |
---|---|
Germanic | Geman, Dutch, Swedish, Danish, Afrikaans, Luxembourgish, Norwegian, Icelandic |
Romance | French, Spanish, Italian, Portuguese, Romanian, Occitan, Asturian, Catalan |
Slavic | Russian, Czech, Polish, Bulgarian, Ukrainian, Serbian, Belarusian, Bosnian |
Indo-Aryan | Hindi, Bengali, Kannada, Marathi, Sindhi, Gujarati, Nepali, Urdu |
Dataset Stats
Resource | Languages | Size |
---|---|---|
High | de, nl, fr, es, ru, cs, hi, bn, ar, he | 5M |
Medium | sv, da, it, pt, pl, bg, kn, mr, mt, ha | 1M |
Low | af, lb, ro, oc, uk, sr, sd, gu, ti, am | 100k |
Extremely-Low | no, is, ast, ca, be, bs, ne, ur, kab, so | 50k |
Build Fairseq dataset (Shard->to avoid RAM OOM)
Read toolkit/build_fairseq_sharded_dataset.sh
Train mTransformer-Large baseline
Read toolkit/train-EC40-mTrans-large.sh
- Downloads last month
- 32