WikiMatrix: Mining 135M Parallel Sentences in 1620 Language Pairs from Wikipedia
This work provides a large-scale parallel corpus for machine translation, especially useful for low-resource and distant language pairs, though it is incremental as it builds on existing multilingual embedding methods.
The authors tackled the problem of automatically extracting parallel sentences from Wikipedia across 85 languages, resulting in a corpus of 135M sentences for 1620 language pairs, with only 34M aligned with English. They trained neural MT systems on this data for 1886 language pairs, achieving strong BLEU scores on the TED corpus, particularly benefiting distant language pairs without English pivoting.
We present an approach based on multilingual sentence embeddings to automatically extract parallel sentences from the content of Wikipedia articles in 85 languages, including several dialects or low-resource languages. We do not limit the the extraction process to alignments with English, but systematically consider all possible language pairs. In total, we are able to extract 135M parallel sentences for 1620 different language pairs, out of which only 34M are aligned with English. This corpus of parallel sentences is freely available at https://github.com/facebookresearch/LASER/tree/master/tasks/WikiMatrix. To get an indication on the quality of the extracted bitexts, we train neural MT baseline systems on the mined data only for 1886 languages pairs, and evaluate them on the TED corpus, achieving strong BLEU scores for many language pairs. The WikiMatrix bitexts seem to be particularly interesting to train MT systems between distant languages without the need to pivot through English.