Towards Language Agnostic Universal Representations

TL;DR

This paper introduces a method to learn language-agnostic representations in machine learning, enabling models trained in one language to perform well in others without additional training, inspired by linguistic theories of universal grammar.

Contribution

The work proposes a novel approach to decouple language from problem representations, allowing zero-shot cross-lingual transfer in machine learning models.

Findings

Models trained on one language with these representations perform similarly in other languages.

The approach is inspired by linguistic theories of universal grammar.

Demonstrates effective zero-shot transfer across languages.

Abstract

When a bilingual student learns to solve word problems in math, we expect the student to be able to solve these problem in both languages the student is fluent in,even if the math lessons were only taught in one language. However, current representations in machine learning are language dependent. In this work, we present a method to decouple the language from the problem by learning language agnostic representations and therefore allowing training a model in one language and applying to a different one in a zero shot fashion. We learn these representations by taking inspiration from linguistics and formalizing Universal Grammar as an optimization process (Chomsky, 2014; Montague, 1970). We demonstrate the capabilities of these representations by showing that the models trained on a single language using language agnostic representations achieve very similar accuracies in other languages.