Cross-lingual language model pretraining

A Conneau, G Lample - Advances in neural information …, 2019‏ - proceedings.neurips.cc
Recent studies have demonstrated the efficiency of generative pretraining for English
natural language understanding. In this work, we extend this approach to multiple …

Massively multilingual sentence embeddings for zero-shot cross-lingual transfer and beyond

M Artetxe, H Schwenk - … of the association for computational linguistics, 2019‏ - direct.mit.edu
We introduce an architecture to learn joint multilingual sentence representations for 93
languages, belonging to more than 30 different families and written in 28 different scripts …

Offline bilingual word vectors, orthogonal transformations and the inverted softmax

SL Smith, DHP Turban, S Hamblin… - arxiv preprint arxiv …, 2017‏ - arxiv.org
Usually bilingual word vectors are trained" online". Mikolov et al. showed they can also be
found" offline", whereby two pre-trained embeddings are aligned with a linear …

SimAlign: High quality word alignments without parallel training data using static and contextualized embeddings

MJ Sabet, P Dufter, F Yvon, H Schütze - arxiv preprint arxiv:2004.08728, 2020‏ - arxiv.org
Word alignments are useful for tasks like statistical and neural machine translation (NMT)
and cross-lingual annotation projection. Statistical word aligners perform well, as do …

From word to sense embeddings: A survey on vector representations of meaning

J Camacho-Collados, MT Pilehvar - Journal of Artificial Intelligence …, 2018‏ - jair.org
Over the past years, distributed semantic representations have proved to be effective and
flexible keepers of prior knowledge to be integrated into downstream applications. This …