Linear Transformations for Cross-lingual Sentiment Analysis

TL;DR

This paper explores zero-shot cross-lingual sentiment analysis across Czech, English, and French using linear transformations with neural classifiers, emphasizing the importance of target domain embeddings for improved performance.

Contribution

It introduces a transformation-based approach for cross-lingual sentiment analysis and compares it with BERT-like models, highlighting the significance of target domain embeddings.

Findings

Target domain embeddings significantly improve cross-lingual classification.

Linear transformations combined with neural classifiers can be effective for cross-lingual tasks.

Transformation-based methods can be competitive with BERT-like models in this context.

Abstract

This paper deals with cross-lingual sentiment analysis in Czech, English and French languages. We perform zero-shot cross-lingual classification using five linear transformations combined with LSTM and CNN based classifiers. We compare the performance of the individual transformations, and in addition, we confront the transformation-based approach with existing state-of-the-art BERT-like models. We show that the pre-trained embeddings from the target domain are crucial to improving the cross-lingual classification results, unlike in the monolingual classification, where the effect is not so distinctive.