Can Code-Switched Texts Activate a Knowledge Switch in LLMs? A Case Study on English-Korean Code-Switching

TL;DR

This paper explores whether code-switching in multilingual texts can activate language-specific knowledge in large language models, especially for low-resource languages, through a new English-Korean question-answering dataset and analysis.

Contribution

It introduces EnKoQA, a synthetic code-switching dataset, and analyzes how code-switching activates knowledge in LLMs for low-resource language tasks.

Findings

Code-switching faithfully activates knowledge in LLMs, especially in language-specific domains.

Multilingual LLMs show improved reasoning with code-switched inputs over monolingual English texts.

Potential of code-switching to enhance low-resource language understanding in LLMs.

Abstract

Recent large language models (LLMs) demonstrate multilingual abilities, yet they are English-centric due to dominance of English in training corpora. The limited resource for low-resource languages remains a crucial challenge. Code-switching (CS), a phenomenon where multilingual speakers alternate between languages in a discourse, can convey subtle cultural and linguistic nuances that can be otherwise lost in translation and elicits language-specific knowledge in human communications. In light of this, we investigate whether code-switching can activate, or identify and leverage knowledge for reasoning when LLMs solve low-resource language tasks. To facilitate the research, we first present EnKoQA, a synthetic English-Korean CS question-answering dataset. We provide comprehensive analysis on a variety of multilingual LLMs by subdividing activation process into knowledge identification and knowledge leveraging. Our results demonstrate that compared to English text, CS can faithfully activate knowledge inside LLMs especially on language-specific domains, suggesting the potential of code-switching on low-resource language tasks.