Language-Specific Sentiment Polarity Biases in Encoder and Large Language Model Classification of Product Reviews
2026-06-22 • Computation and Language
Computation and Language
AI summaryⓘ
The authors studied how AI models understand positive and negative feelings in reviews written in different languages. They found that large language models are better at spotting negative reviews in French but tend to miss indirect criticism in Japanese, where encoder models do better on positive reviews. These findings mean that AI might not be equally good at reading emotions in every language, which is important for businesses and social studies using these tools. The authors highlight the need to consider language and model type when working with sentiment analysis.
sentiment analysispolarity biaslarge language modelsencoder modelsFrench languageJapanese languageindirect criticismmultilingual AInatural language processingmodel accuracy
Authors
Advita Rajiv, Kavitha Kothur, Gautham Reddy
Abstract
This study investigates sentiment polarity biases, specifically, differences in how accurately AI models classify positive versus negative reviews across languages and model architectures. Large language models show a negative bias in French and are more accurate on negative reviews, while encoder models exhibit positive bias in Japanese, missing negative reviews that use indirect criticism. These language-specific polarity biases have implications in both social and business domains deploying multilingual sentiment analysis systems.