A robust self-learning method for fully unsupervised cross-lingual mappings of word embeddings

TL;DR

This paper introduces a fully unsupervised, robust self-learning method for cross-lingual word embedding mappings that outperforms previous supervised and unsupervised approaches, especially in realistic scenarios.

Contribution

It presents a novel initialization and self-learning algorithm that reliably improves cross-lingual embeddings without parallel data, outperforming prior methods.

Findings

Successfully applied in diverse scenarios

Achieves state-of-the-art results on standard datasets

Surpasses previous supervised systems

Abstract

Recent work has managed to learn cross-lingual word embeddings without parallel data by mapping monolingual embeddings to a shared space through adversarial training. However, their evaluation has focused on favorable conditions, using comparable corpora or closely-related languages, and we show that they often fail in more realistic scenarios. This work proposes an alternative approach based on a fully unsupervised initialization that explicitly exploits the structural similarity of the embeddings, and a robust self-learning algorithm that iteratively improves this solution. Our method succeeds in all tested scenarios and obtains the best published results in standard datasets, even surpassing previous supervised systems. Our implementation is released as an open source project at https://github.com/artetxem/vecmap