Examining the Joint Influence of Sarcasm and Emotion on Gender Prediction

The use of sarcasm and the expression of emotions can be seen as key differences in the communication styles often associated with different genders. The relationship between gender, sarcasm, and emotion in online communication is explored in this study. Automatic gender identification is crucial for various services relying on user background data. However, due to privacy concerns, gender information is often unavailable for some social media users. Leveraging the differences in communication styles between genders, this study proposes a neural network based approach to automatically detect users’ gender from their textual posts on social media platforms. We analyze the effects of emotion and sarcasm on gender prediction, employing neural network-based architectures. Our model extracts tweet features using bidirectional gated recurrent units. Sarcasm intensity is integrated with tweet representations. Finally, the gender labels are predicted at the classification layer. Experiments are conducted using the PAN-2018 dataset, demonstrating the comparative performance of our models compared to previous works. Additionally, we investigate the impact of incorporating emotion and sarcasm information on gender prediction across different models.