Reducing social biases in text-based emotion prediction using semantic blinding and semantic propagation graph neural networks

Abstract

Abstract This paper introduces the Semantic Propagation Graph Neural Network (SProp GNN), a machine learning emotion prediction (EP) architecture that relies exclusively on syntactic structures and word-level emotional cues to predict emotions in text. By semantically blinding the model to information about specific words, it is robust to social biases such as political or gender bias that have been plaguing previous machine learning-based EP systems. The SProp GNN shows performance superior to lexicon-based alternatives such as VADER (Valence Aware Dictionary and Sentiment Reasoner) and Emotlas on two different prediction tasks, and across two languages. Additionally, it approaches the accuracy of transformer-based models – with an average of 5.7% difference in accuracy scores on discrete English benchmarks, and from 0.075 to 0.145 correlation difference on English and Polish dimensional emotion prediction datasets - while significantly reducing bias in emotion prediction tasks. By offering improved explainability and reducing bias, the SProp GNN bridges the methodological gap between interpretable lexicon approaches and powerful, yet often opaque, deep learning models, offering a robust tool for fair and effective emotion prediction in understanding human behavior through text.