Mono- / multilingual

• Making Monolingual Sentence Embeddings Multilingual using Knowledge Distillation, by Nils Reimers, Iryna Gurevych [ paper ] An easy and efficient method to extend existing sentence embedding models to new languages. This allows to create multilingual versions from previously monolingual models.

• Multilingual Universal Sentence Encoder for Semantic Retrieval, by Yinfei Yang, Daniel Cer, Amin Ahmad, Mandy Guo, Jax Law, Noah Constant, Gustavo Hernandez Abrego, Steve Yuan, Chris Tar, Yun-Hsuan Sung, Brian Strope, Ray Kurzweil [ paper ] When the authors team is from Google and it comprises of 12 people, it's a sign of something worth reading. - They propose two pre-trained multilingual (16 languages) retrieval models based on Transformer and CNN model architectures. - They reach SOTA performance on the folowing tasks: semantic retrieval (SR), translation pair bitext retrieval (BR) and retrieval question answering (ReQA). - They make the models available on TensorFlow Hub. - They also wrote a short blog post about their findngs on Google Blog.