The MUSE dataset contains bilingual dictionaries for 110 pairs of languages. For each language pair, the training seed dictionaries contain approximately 5000 word pairs while the evaluation sets contain 1500 word pairs.
2 PAPERS • 2 BENCHMARKS