On the Applicability of Zero-Shot Cross-Lingual Transfer Learning for Sentiment Classification in Distant Language Pairs

TL;DR

This paper evaluates the effectiveness of zero-shot cross-lingual transfer learning with XLM-R for sentiment classification in Japanese and Indonesian, demonstrating competitive results without target language training.

Contribution

It provides an empirical assessment of XLM-R's zero-shot transfer capabilities across Japanese and Indonesian, highlighting the potential of multi-lingual models for sentiment analysis.

Findings

Achieved best results on one Japanese dataset

Comparable results across other datasets in Japanese and Indonesian

Supports training of a single multi-lingual model

Abstract

This research explores the applicability of cross-lingual transfer learning from English to Japanese and Indonesian using the XLM-R pre-trained model. The results are compared with several previous works, either by models using a similar zero-shot approach or a fully-supervised approach, to provide an overview of the zero-shot transfer learning approach's capability using XLM-R in comparison with existing models. Our models achieve the best result in one Japanese dataset and comparable results in other datasets in Japanese and Indonesian languages without being trained using the target language. Furthermore, the results suggest that it is possible to train a multi-lingual model, instead of one model for each language, and achieve promising results.