Making Monolingual Sentence Embeddings Multilingual using Knowledge Distillation

Nils Reimers, Iryna Gurevych


Abstract
We present an easy and efficient method to extend existing sentence embedding models to new languages. This allows to create multilingual versions from previously monolingual models. The training is based on the idea that a translated sentence should be mapped to the same location in the vector space as the original sentence. We use the original (monolingual) model to generate sentence embeddings for the source language and then train a new system on translated sentences to mimic the original model. Compared to other methods for training multilingual sentence embeddings, this approach has several advantages: It is easy to extend existing models with relatively few samples to new languages, it is easier to ensure desired properties for the vector space, and the hardware requirements for training are lower. We demonstrate the effectiveness of our approach for 50+ languages from various language families. Code to extend sentence embeddings models to more than 400 languages is publicly available.
Anthology ID:
2020.emnlp-main.365
Volume:
Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP)
Month:
November
Year:
2020
Address:
Online
Venue:
EMNLP
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
4512–4525
Language:
URL:
https://aclanthology.org/2020.emnlp-main.365
DOI:
10.18653/v1/2020.emnlp-main.365
Bibkey:
Copy Citation:
PDF:
https://aclanthology.org/2020.emnlp-main.365.pdf
Video:
 https://slideslive.com/38938673