Dual-Alignment Pre-training for Cross-lingual Sentence Embedding

TL;DR

This paper introduces a dual-alignment pre-training framework that enhances cross-lingual sentence embeddings by combining sentence-level and token-level alignment, using a novel representation translation learning task.

Contribution

It proposes a new dual-alignment pre-training method with a representation translation learning task for better multilingual sentence embeddings.

Findings

Significant improvements on three cross-lingual benchmarks

RTL is more suitable and efficient than translation language modeling

Effective integration of token-level and sentence-level alignment

Abstract

Recent studies have shown that dual encoder models trained with the sentence-level translation ranking task are effective methods for cross-lingual sentence embedding. However, our research indicates that token-level alignment is also crucial in multilingual scenarios, which has not been fully explored previously. Based on our findings, we propose a dual-alignment pre-training (DAP) framework for cross-lingual sentence embedding that incorporates both sentence-level and token-level alignment. To achieve this, we introduce a novel representation translation learning (RTL) task, where the model learns to use one-side contextualized token representation to reconstruct its translation counterpart. This reconstruction objective encourages the model to embed translation information into the token representation. Compared to other token-level alignment methods such as translation language modeling, RTL is more suitable for dual encoder architectures and is computationally efficient. Extensive experiments on three sentence-level cross-lingual benchmarks demonstrate that our approach can significantly improve sentence embedding. Our code is available at https://github.com/ChillingDream/DAP.