Improved Sentiment Detection via Label Transfer from Monolingual to Synthetic Code-Switched Text

TL;DR

This paper introduces a method to generate synthetic code-switched text from monolingual data, significantly improving sentiment and hate speech detection accuracy in multilingual contexts involving minority languages.

Contribution

The authors propose a novel technique for synthesizing labeled code-switched text by replacing subtrees in parse trees with translated token spans, enhancing sentiment analysis models.

Findings

Sentiment detection accuracy improved by up to 7.20% across three language pairs.

Hate speech detection accuracy increased by 4-6% with synthetic and augmented data.

Synthetic data augmentation yields significant performance gains in multilingual sentiment tasks.

Abstract

Multilingual writers and speakers often alternate between two languages in a single discourse, a practice called "code-switching". Existing sentiment detection methods are usually trained on sentiment-labeled monolingual text. Manually labeled code-switched text, especially involving minority languages, is extremely rare. Consequently, the best monolingual methods perform relatively poorly on code-switched text. We present an effective technique for synthesizing labeled code-switched text from labeled monolingual text, which is more readily available. The idea is to replace carefully selected subtrees of constituency parses of sentences in the resource-rich language with suitable token spans selected from automatic translations to the resource-poor language. By augmenting scarce human-labeled code-switched text with plentiful synthetic code-switched text, we achieve significant improvements in sentiment labeling accuracy (1.5%, 5.11%, 7.20%) for three different language pairs (English-Hindi, English-Spanish and English-Bengali). We also get significant gains for hate speech detection: 4% improvement using only synthetic text and 6% if augmented with real text.