# WikiMatrix: Mining 135M Parallel Sentences in 1620 Language Pairs from Wikipedia

## TL;DR

This paper introduces WikiMatrix, a large multilingual parallel corpus extracted from Wikipedia using sentence embeddings, enabling improved machine translation for many language pairs, especially low-resource and distant languages.

## Contribution

It presents a novel method for mining a vast multilingual parallel corpus from Wikipedia without limiting to English, covering 1620 language pairs with 135 million sentences.

## Key findings

- Successfully extracted 135 million parallel sentences across 1620 language pairs.
- Neural MT systems trained on WikiMatrix data achieve strong BLEU scores.
- The corpus is especially beneficial for translating distant and low-resource language pairs.

## Abstract

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.