Lingua Custodi's participation at the WMT 2025 Terminology shared task

TL;DR

This paper explores multilingual sentence embeddings using BERT-based models, demonstrating significant reductions in training data needs and achieving state-of-the-art retrieval accuracy across 112 languages, with practical applications in translation.

Contribution

It introduces a novel multilingual sentence embedding approach that combines multiple methods, significantly reducing data requirements and outperforming existing models like LASER.

Findings

Achieves 83.7% bi-text retrieval accuracy over 112 languages.

Reduces parallel training data needs by 80%.

Enables training competitive NMT models for en-zh and en-de.

Abstract

While BERT is an effective method for learning monolingual sentence embeddings for semantic similarity and embedding based transfer learning BERT based cross-lingual sentence embeddings have yet to be explored. We systematically investigate methods for learning multilingual sentence embeddings by combining the best methods for learning monolingual and cross-lingual representations including: masked language modeling (MLM), translation language modeling (TLM), dual encoder translation ranking, and additive margin softmax. We show that introducing a pre-trained multilingual language model dramatically reduces the amount of parallel training data required to achieve good performance by 80%. Composing the best of these methods produces a model that achieves 83.7% bi-text retrieval accuracy over 112 languages on Tatoeba, well above the 65.5 achieved by LASER, while still performing competitively on monolingual transfer learning benchmarks. Parallel data mined from CommonCrawl using our best model is shown to train competitive NMT models for en-zh and en-de. We publicly release our best multilingual sentence embedding model for 109+ languages at https://tfhub.dev/google/LaBSE.