A compact multilingual embedding model for efficient cross-lingual retrieval and sentence similarity.