Zero-Shot Cross-Lingual NER Using Phonemic Representations for Low-Resource Languages

TL;DR

This paper introduces a phonemic representation-based method for zero-shot cross-lingual NER that effectively handles low-resource languages and non-Latin scripts, outperforming baseline models.

Contribution

It presents a novel phonemic approach using IPA for zero-shot NER, reducing reliance on prior language knowledge and improving performance in low-resource scenarios.

Findings

Achieved highest average F1 score of 46.38% in low-resource languages

Significantly outperformed baseline models in non-Latin scripts

Demonstrated robustness with low-resource and diverse language scripts

Abstract

Existing zero-shot cross-lingual NER approaches require substantial prior knowledge of the target language, which is impractical for low-resource languages. In this paper, we propose a novel approach to NER using phonemic representation based on the International Phonetic Alphabet (IPA) to bridge the gap between representations of different languages. Our experiments show that our method significantly outperforms baseline models in extremely low-resource languages, with the highest average F1 score (46.38%) and lowest standard deviation (12.67), particularly demonstrating its robustness with non-Latin scripts. Our codes are available at https://github.com/Gabriel819/zeroshot_ner.git