Improving the Generalizability of Text-Based Emotion Detection by Leveraging Transformers with Psycholinguistic Features
This work addresses the challenge of deploying emotion detection models in real-world applications by enhancing their robustness across different domains, though it appears incremental as it builds on existing transformer and feature-based methods.
The paper tackled the problem of poor out-of-domain generalizability in text-based emotion detection by proposing hybrid models combining transformers with psycholinguistic features and BiLSTM networks, resulting in improved generalization to out-of-distribution data and competitive in-domain performance.
In recent years, there has been increased interest in building predictive models that harness natural language processing and machine learning techniques to detect emotions from various text sources, including social media posts, micro-blogs or news articles. Yet, deployment of such models in real-world sentiment and emotion applications faces challenges, in particular poor out-of-domain generalizability. This is likely due to domain-specific differences (e.g., topics, communicative goals, and annotation schemes) that make transfer between different models of emotion recognition difficult. In this work we propose approaches for text-based emotion detection that leverage transformer models (BERT and RoBERTa) in combination with Bidirectional Long Short-Term Memory (BiLSTM) networks trained on a comprehensive set of psycholinguistic features. First, we evaluate the performance of our models within-domain on two benchmark datasets: GoEmotion and ISEAR. Second, we conduct transfer learning experiments on six datasets from the Unified Emotion Dataset to evaluate their out-of-domain robustness. We find that the proposed hybrid models improve the ability to generalize to out-of-distribution data compared to a standard transformer-based approach. Moreover, we observe that these models perform competitively on in-domain data.