Since the Scientific Literature Is Multilingual, Our Models Should Be Too

Domaine:

natural language processing

Type de record:

paper

Créateur:

EbrChu

Hôte:

English has long been assumed the $\textit{lingua franca}$ of scientific research, and this notion is reflected in the natural language processing (NLP) research involving scientific document representation. In this position piece, we quantitatively show that the literature is largely multilingual and argue that current models and benchmarks should reflect this linguistic diversity. We provide evidence that text-based models fail to create meaningful representations for non-English papers and highlight the negative user-facing impacts of using English-only models non-discriminately across a multilingual domain. We end with suggestions for the NLP community on how to improve performance on non-English documents.

Visit

arxiv.org

Similaires

Should LLMs be WEIRD? Exploring WEIRDness and Human Rights in Large Language Models“What African Literature is in the Complexity of The Interpreters (Wole Soyinka)? Is there any way our Biased Ugandan Readership could be Re-educated to Like its Style and Appreciate its Message?”Should Xhosa male initiation be abolished?Which Immigrants Should Be Naturalized? Which Should Be Deported? Evidence from a Survey Experiment in Côte d’IvoireNil-Filing in Eswatini: Should the Revenue Administration be Concerned?How soon should patients be eligible for differentiated service delivery models for antiretroviral treatment? Evidence from Zambia

Should LLMs be WEIRD? Exploring WEIRDness and Human Rights in Large Language Models

Large language models (LLMs) are often trained on data that reflect WEIRD values: Western, Educated,

“What African Literature is in the Complexity of The Interpreters (Wole Soyinka)? Is there any way our Biased Ugandan Readership could be Re-educated to Like its Style and Appreciate its Message?”

This article re-examines views of some literary critics who find Soyinka’s style intolerably complex

Should Xhosa male initiation be abolished?

The Xhosa male initiation practice, as it was viewed in the past, is discussed by focusing on its im

Which Immigrants Should Be Naturalized? Which Should Be Deported? Evidence from a Survey Experiment in Côte d’Ivoire

Experimental studies on immigration attitudes have been conducted overwhelmingly in Western countrie

Nil-Filing in Eswatini: Should the Revenue Administration be Concerned?

The tax-to-GDP ratio in developing countries is still remarkably low for many different reasons. One

How soon should patients be eligible for differentiated service delivery models for antiretroviral treatment? Evidence from Zambia

Abstract Introduction Attrition from HIV t