Relationship of the language distance to English ability of a country

  • 2022-11-15 02:40:00
  • Cao Xinxin, Lei Xiaolan, Murtadha Ahmed
  • 2

Abstract

Language difference is one of the factors that hinder the acquisition ofsecond language skills. In this article, we introduce a novel solution thatleverages the strength of deep neural networks to measure the semanticdissimilarity between languages based on their word distributions in theembedding space of the multilingual pre-trained language model (e.g.,BERT).Then, we empirically examine the effectiveness of the proposed semanticlanguage distance (SLD) in explaining the consistent variation in Englishability of countries, which is proxied by their performance in theInternet-Based Test of English as Foreign Language (TOEFL iBT). Theexperimental results show that the language distance demonstrates negativeinfluence on a country's average English ability. Interestingly, the effect ismore significant on speaking and writing subskills, which pertain to theproductive aspects of language learning. Besides, we provide specificrecommendations for future research directions.

 

Quick Read (beta)

loading the full paper ...