A Simple and Effective Method To Eliminate the Self Language Bias in Multilingual Representations

TL;DR

This paper introduces a simple, post-training linear method called LIR to remove language bias from multilingual models, significantly improving cross-lingual tasks by isolating semantic content from language identity.

Contribution

The paper proposes a novel, model-agnostic linear approach using geometric algebra to eliminate language bias in multilingual representations, enhancing cross-lingual transfer.

Findings

LIR achieves nearly 100% relative improvement in MAP on LAReQA for weak-alignment models.

Removing language information improves cross-lingual transfer performance.

LIR is simple, effective, and applicable post-training without model modifications.

Abstract

Language agnostic and semantic-language information isolation is an emerging research direction for multilingual representations models. We explore this problem from a novel angle of geometric algebra and semantic space. A simple but highly effective method "Language Information Removal (LIR)" factors out language identity information from semantic related components in multilingual representations pre-trained on multi-monolingual data. A post-training and model-agnostic method, LIR only uses simple linear operations, e.g. matrix factorization and orthogonal projection. LIR reveals that for weak-alignment multilingual systems, the principal components of semantic spaces primarily encodes language identity information. We first evaluate the LIR on a cross-lingual question answer retrieval task (LAReQA), which requires the strong alignment for the multilingual embedding space. Experiment shows that LIR is highly effectively on this task, yielding almost 100% relative improvement in MAP for weak-alignment models. We then evaluate the LIR on Amazon Reviews and XEVAL dataset, with the observation that removing language information is able to improve the cross-lingual transfer performance.