Enhancing Emotion Recognition in Text with Stacked CNN-BiLSTM Framework

Emotions play a significant impact on shaping the human experience, and humans tend to express emotions in all forms of communication, including text, especially in the social media era. Recognizing emotions in text is crucial for various applications, enabling a better understanding and the design of emotion-aware responses. This study proposes a novel framework for textual emotion recognition using a stacked CNN-BiLSTM. This framework aims to leverage the advantages of both the algorithms effectively, especially in the context of the increasingly intricate conversation found in social media. Further, the efficacy of the framework with hybrid embedding methods is demonstrated in the experiments using various optimizers and activation functions. The promising results validate the application of this novel approach for inherently complex input texts.